cascade-ai 0.48.0 → 0.49.0

package/dist/index.js

@@ -192,8 +192,1660 @@ var init_audit_logger = __esm({
   }
 });
 
+// src/core/router/pricing-data.json
+var pricing_data_default = {
+  entries: [
+    {
+      model: "claude-opus-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 5,
+          outputPer1m: 25,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25,
+          reasoningOutputPer1m: 25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-opus-4-8",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 5,
+          outputPer1m: 25,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25,
+          reasoningOutputPer1m: 25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-opus-4-7",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 5,
+          outputPer1m: 25,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25,
+          reasoningOutputPer1m: 25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-opus-4-6",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 5,
+          outputPer1m: 25,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25,
+          reasoningOutputPer1m: 25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-opus-4-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 5,
+          outputPer1m: 25,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25,
+          reasoningOutputPer1m: 25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5
+    },
+    {
+      model: "claude-opus-4-1",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 15,
+          outputPer1m: 75,
+          cachedInputPer1m: 1.5,
+          cacheWritePer1m: 18.75,
+          reasoningOutputPer1m: 75
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5,
+      notes: "Deprecated by Anthropic; priced for historical/cost-reporting accuracy."
+    },
+    {
+      model: "claude-opus-4",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 15,
+          outputPer1m: 75,
+          cachedInputPer1m: 1.5,
+          cacheWritePer1m: 18.75,
+          reasoningOutputPer1m: 75
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5,
+      notes: "Retired except on Google Cloud; priced for historical/cost-reporting accuracy."
+    },
+    {
+      model: "claude-sonnet-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 2,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.2,
+          cacheWritePer1m: 2.5,
+          reasoningOutputPer1m: 10
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6,
+      notes: "Introductory pricing through 2026-08-31. Standard pricing of $3/$15 per 1M takes effect 2026-09-01 \u2014 re-check this entry then."
+    },
+    {
+      model: "claude-sonnet-4-6",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 3,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.3,
+          cacheWritePer1m: 3.75,
+          reasoningOutputPer1m: 15
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-sonnet-4-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 3,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.3,
+          cacheWritePer1m: 3.75,
+          reasoningOutputPer1m: 15
+        },
+        {
+          inputPer1m: 6,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.6,
+          cacheWritePer1m: 7.5,
+          reasoningOutputPer1m: 22.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5,
+      notes: "Long-context (>200K) premium applies to the 1M-context beta. Base rates from the Anthropic pricing page; the >200K tier from the litellm catalog (Anthropic does not tabulate it on the pricing page)."
+    },
+    {
+      model: "claude-sonnet-4",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 3,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.3,
+          cacheWritePer1m: 3.75,
+          reasoningOutputPer1m: 15
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5
+    },
+    {
+      model: "claude-haiku-4-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 1,
+          outputPer1m: 5,
+          cachedInputPer1m: 0.1,
+          cacheWritePer1m: 1.25,
+          reasoningOutputPer1m: 5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5
+    },
+    {
+      model: "claude-haiku-3-5",
+      provider: "anthropic",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://platform.claude.com/docs/en/about-claude/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 0.8,
+          outputPer1m: 4,
+          cachedInputPer1m: 0.08,
+          cacheWritePer1m: 1,
+          reasoningOutputPer1m: 4
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 2e5
+    },
+    {
+      model: "gpt-5.6-sol",
+      provider: "openai",
+      aliases: [
+        "gpt-5.6"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 5,
+          outputPer1m: 30,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25
+        },
+        {
+          inputPer1m: 10,
+          outputPer1m: 45,
+          cachedInputPer1m: 1,
+          cacheWritePer1m: 12.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.6-terra",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25,
+          cacheWritePer1m: 3.125
+        },
+        {
+          inputPer1m: 5,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.5,
+          cacheWritePer1m: 6.25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.6-luna",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 1,
+          outputPer1m: 6,
+          cachedInputPer1m: 0.1,
+          cacheWritePer1m: 1.25
+        },
+        {
+          inputPer1m: 2,
+          outputPer1m: 9,
+          cachedInputPer1m: 0.2,
+          cacheWritePer1m: 2.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.5",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 5,
+          outputPer1m: 30,
+          cachedInputPer1m: 0.5
+        },
+        {
+          inputPer1m: 10,
+          outputPer1m: 45,
+          cachedInputPer1m: 1
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.4",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25
+        },
+        {
+          inputPer1m: 5,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.4-mini",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.75,
+          outputPer1m: 4.5,
+          cachedInputPer1m: 0.075
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 1.25,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.125
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-5-mini",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.25,
+          outputPer1m: 2,
+          cachedInputPer1m: 0.025
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-5-nano",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.05,
+          outputPer1m: 0.4,
+          cachedInputPer1m: 5e-3
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-4.1",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2,
+          outputPer1m: 8,
+          cachedInputPer1m: 0.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1047576
+    },
+    {
+      model: "gpt-4.1-mini",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.4,
+          outputPer1m: 1.6,
+          cachedInputPer1m: 0.1
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1047576
+    },
+    {
+      model: "gpt-4.1-nano",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.1,
+          outputPer1m: 0.4,
+          cachedInputPer1m: 0.025
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1047576
+    },
+    {
+      model: "gpt-4o",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2.5,
+          outputPer1m: 10,
+          cachedInputPer1m: 1.25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 128e3
+    },
+    {
+      model: "gpt-4o-mini",
+      provider: "openai",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.15,
+          outputPer1m: 0.6,
+          cachedInputPer1m: 0.075
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 128e3
+    },
+    {
+      model: "gpt-5.5",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 5,
+          outputPer1m: 30,
+          cachedInputPer1m: 0.5
+        },
+        {
+          inputPer1m: 10,
+          outputPer1m: 45,
+          cachedInputPer1m: 1
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.5",
+      provider: "azure",
+      region: "us",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 5.5,
+          outputPer1m: 33,
+          cachedInputPer1m: 0.55
+        },
+        {
+          inputPer1m: 11,
+          outputPer1m: 49.5,
+          cachedInputPer1m: 1.1
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.5",
+      provider: "azure",
+      region: "eu",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 5.5,
+          outputPer1m: 33,
+          cachedInputPer1m: 0.55
+        },
+        {
+          inputPer1m: 11,
+          outputPer1m: 49.5,
+          cachedInputPer1m: 1.1
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.4",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25
+        },
+        {
+          inputPer1m: 5,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.5
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.4",
+      provider: "azure",
+      region: "us",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2.75,
+          outputPer1m: 16.5,
+          cachedInputPer1m: 0.28
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.4",
+      provider: "azure",
+      region: "eu",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2.75,
+          outputPer1m: 16.5,
+          cachedInputPer1m: 0.28
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.6-terra",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25
+        },
+        {
+          inputPer1m: 5,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.5
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5.6-terra",
+      provider: "azure",
+      region: "eu",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 272e3,
+          inputPer1m: 2.75,
+          outputPer1m: 16.5,
+          cachedInputPer1m: 0.275
+        },
+        {
+          inputPer1m: 5.5,
+          outputPer1m: 24.75,
+          cachedInputPer1m: 0.55
+        }
+      ],
+      contextWindow: 105e4
+    },
+    {
+      model: "gpt-5",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 1.25,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.125
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-5-mini",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.25,
+          outputPer1m: 2,
+          cachedInputPer1m: 0.025
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-5-nano",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.05,
+          outputPer1m: 0.4,
+          cachedInputPer1m: 5e-3
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-4.1",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2,
+          outputPer1m: 8,
+          cachedInputPer1m: 0.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1047576
+    },
+    {
+      model: "gpt-4o",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 2.5,
+          outputPer1m: 10,
+          cachedInputPer1m: 1.25
+        }
+      ],
+      contextWindow: 128e3
+    },
+    {
+      model: "gpt-4o-mini",
+      provider: "azure",
+      region: "global",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.165,
+          outputPer1m: 0.66,
+          cachedInputPer1m: 0.075
+        }
+      ],
+      contextWindow: 128e3,
+      notes: "Azure charges $0.165/$0.66 per 1M for gpt-4o-mini vs $0.15/$0.60 on OpenAI \u2014 same model, different price."
+    },
+    {
+      model: "gemini-3.1-pro",
+      provider: "gemini",
+      aliases: [
+        "gemini-3.1-pro-preview"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 2,
+          outputPer1m: 12,
+          cachedInputPer1m: 0.2,
+          reasoningOutputPer1m: 12
+        },
+        {
+          inputPer1m: 4,
+          outputPer1m: 18,
+          cachedInputPer1m: 0.4,
+          reasoningOutputPer1m: 18
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576,
+      notes: "Long-context tier: $2/$12 per 1M up to 200K prompt tokens, $4/$18 above it."
+    },
+    {
+      model: "gemini-3-pro",
+      provider: "gemini",
+      aliases: [
+        "gemini-3-pro-preview"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 2,
+          outputPer1m: 12,
+          cachedInputPer1m: 0.2,
+          reasoningOutputPer1m: 12
+        },
+        {
+          inputPer1m: 4,
+          outputPer1m: 18,
+          cachedInputPer1m: 0.4,
+          reasoningOutputPer1m: 18
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-3.6-flash",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 1.5,
+          outputPer1m: 7.5,
+          cachedInputPer1m: 0.15,
+          reasoningOutputPer1m: 7.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576,
+      notes: "Released 2026-07-21. Output pricing dropped ~17% from Gemini 3.5 Flash ($9.00 -> $7.50 per 1M); input unchanged."
+    },
+    {
+      model: "gemini-3.5-flash",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 1.5,
+          outputPer1m: 9,
+          cachedInputPer1m: 0.15,
+          reasoningOutputPer1m: 9
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-3.5-flash-lite",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/pricing",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.3,
+          outputPer1m: 2.5,
+          cachedInputPer1m: 0.03,
+          reasoningOutputPer1m: 2.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-3.1-flash-lite",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/pricing#gemini-3.1-flash-lite",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.25,
+          outputPer1m: 1.5,
+          cachedInputPer1m: 0.025,
+          reasoningOutputPer1m: 1.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-3-flash",
+      provider: "gemini",
+      aliases: [
+        "gemini-3-flash-preview"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/pricing/gemini-3",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.5,
+          outputPer1m: 3,
+          cachedInputPer1m: 0.05,
+          reasoningOutputPer1m: 3
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.5-pro",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 1.25,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.125
+        },
+        {
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.5-flash",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.3,
+          outputPer1m: 2.5,
+          cachedInputPer1m: 0.03,
+          reasoningOutputPer1m: 2.5
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.5-flash-lite",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-lite",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.1,
+          outputPer1m: 0.4,
+          cachedInputPer1m: 0.01,
+          reasoningOutputPer1m: 0.4
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.0-flash",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/pricing#2_0flash",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.1,
+          outputPer1m: 0.4,
+          cachedInputPer1m: 0.025
+        }
+      ],
+      contextWindow: 1048576,
+      notes: "Gemini Developer API price. Vertex AI lists $0.15/$0.60 per 1M for the same model \u2014 see the vertex entry."
+    },
+    {
+      model: "gemini-2.0-flash-lite",
+      provider: "gemini",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/pricing#gemini-2.0-flash-lite",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.075,
+          outputPer1m: 0.3,
+          cachedInputPer1m: 0.01875
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.0-flash",
+      provider: "vertex",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 0.15,
+          outputPer1m: 0.6
+        }
+      ],
+      batchMultiplier: 0.5,
+      notes: "Vertex AI lists $0.15/$0.60 per 1M; the Gemini Developer API lists $0.10/$0.40 for the same model."
+    },
+    {
+      model: "gemini-3.1-pro",
+      provider: "vertex",
+      aliases: [
+        "gemini-3.1-pro-preview"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 2,
+          outputPer1m: 12,
+          cachedInputPer1m: 0.2,
+          reasoningOutputPer1m: 12
+        },
+        {
+          inputPer1m: 4,
+          outputPer1m: 18,
+          cachedInputPer1m: 0.4,
+          reasoningOutputPer1m: 18
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-3.6-flash",
+      provider: "vertex",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 1.5,
+          outputPer1m: 7.5,
+          cachedInputPer1m: 0.15,
+          reasoningOutputPer1m: 7.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.5-pro",
+      provider: "vertex",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 1.25,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.13
+        },
+        {
+          inputPer1m: 2.5,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.25
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "gemini-2.5-flash",
+      provider: "vertex",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "vendor-pricing-page",
+      tiers: [
+        {
+          inputPer1m: 0.3,
+          outputPer1m: 2.5,
+          cachedInputPer1m: 0.03,
+          reasoningOutputPer1m: 2.5
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 1048576
+    },
+    {
+      model: "claude-sonnet-4-6",
+      provider: "openrouter",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://openrouter.ai/anthropic/claude-sonnet-4.6",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 3,
+          outputPer1m: 15,
+          cachedInputPer1m: 0.3,
+          cacheWritePer1m: 3.75
+        },
+        {
+          inputPer1m: 6,
+          outputPer1m: 22.5,
+          cachedInputPer1m: 0.6,
+          cacheWritePer1m: 7.5
+        }
+      ],
+      contextWindow: 1e6
+    },
+    {
+      model: "claude-haiku-4-5",
+      provider: "openrouter",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 1,
+          outputPer1m: 5,
+          cachedInputPer1m: 0.1,
+          cacheWritePer1m: 1.25
+        }
+      ],
+      contextWindow: 2e5
+    },
+    {
+      model: "gpt-5",
+      provider: "openrouter",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 1.25,
+          outputPer1m: 10,
+          cachedInputPer1m: 0.125
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gpt-5-mini",
+      provider: "openrouter",
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.25,
+          outputPer1m: 2,
+          cachedInputPer1m: 0.025
+        }
+      ],
+      contextWindow: 272e3
+    },
+    {
+      model: "gemini-3.1-pro",
+      provider: "openrouter",
+      aliases: [
+        "gemini-3.1-pro-preview"
+      ],
+      modality: "text",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://openrouter.ai/google/gemini-3.1-pro-preview",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          maxInputTokens: 2e5,
+          inputPer1m: 2,
+          outputPer1m: 12,
+          cachedInputPer1m: 0.2
+        },
+        {
+          inputPer1m: 4,
+          outputPer1m: 18,
+          cachedInputPer1m: 0.4,
+          cacheWritePer1m: 0.25
+        }
+      ],
+      contextWindow: 1048576
+    },
+    {
+      model: "text-embedding-3-small",
+      provider: "openai",
+      modality: "embedding",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.02,
+          outputPer1m: 0
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 8191
+    },
+    {
+      model: "text-embedding-3-large",
+      provider: "openai",
+      modality: "embedding",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.13,
+          outputPer1m: 0
+        }
+      ],
+      batchMultiplier: 0.5,
+      contextWindow: 8191
+    },
+    {
+      model: "gemini-embedding-001",
+      provider: "gemini",
+      modality: "embedding",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/embeddings#model-versions",
+      provenance: "litellm-catalog",
+      tiers: [
+        {
+          inputPer1m: 0.15,
+          outputPer1m: 0
+        }
+      ],
+      contextWindow: 2048
+    },
+    {
+      model: "dall-e-3",
+      provider: "openai",
+      modality: "image",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_image",
+          amount: 0.04,
+          variant: "1024x1024 standard"
+        }
+      ]
+    },
+    {
+      model: "gemini-3-pro-image",
+      provider: "gemini",
+      modality: "image",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/pricing",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_image",
+          amount: 0.134,
+          variant: "default output resolution"
+        }
+      ],
+      notes: "Google bills image output per output-token ($120 per 1M image output tokens); $0.134/image is that rate applied to one default-resolution image. Cost per image therefore varies with resolution."
+    },
+    {
+      model: "imagen-4.0-generate-001",
+      provider: "vertex",
+      modality: "image",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/pricing",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_image",
+          amount: 0.04
+        }
+      ]
+    },
+    {
+      model: "veo-3.1-generate-001",
+      provider: "gemini",
+      modality: "video",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://ai.google.dev/gemini-api/docs/video",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_second",
+          amount: 0.4
+        }
+      ]
+    },
+    {
+      model: "veo-3.1-generate-001",
+      provider: "vertex",
+      modality: "video",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/veo",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_second",
+          amount: 0.4
+        }
+      ]
+    },
+    {
+      model: "tts-1",
+      provider: "openai",
+      modality: "speech",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_character",
+          amount: 15e-6
+        }
+      ],
+      notes: "$15 per 1M characters."
+    },
+    {
+      model: "tts-1-hd",
+      provider: "openai",
+      modality: "speech",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_character",
+          amount: 3e-5
+        }
+      ],
+      notes: "$30 per 1M characters."
+    },
+    {
+      model: "whisper-1",
+      provider: "openai",
+      modality: "transcription",
+      currency: "USD",
+      asOf: "2026-07-27",
+      source: "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json",
+      provenance: "litellm-catalog",
+      rates: [
+        {
+          unit: "per_audio_minute",
+          amount: 6e-3
+        }
+      ],
+      notes: "litellm records $0.0001 per audio second; $0.006/minute is that rate x 60."
+    }
+  ]
+};
+
+// src/core/router/pricing.ts
+var DATASET = pricing_data_default;
+function normalizeForPricing(id) {
+  let s = id.trim().toLowerCase();
+  s = s.replace(/^models\//, "");
+  const slash = s.lastIndexOf("/");
+  if (slash !== -1) s = s.slice(slash + 1);
+  const tag = s.search(/[:@]/);
+  if (tag !== -1) s = s.slice(0, tag);
+  s = s.replace(/-latest$/, "");
+  s = s.replace(/-preview(?:-\d{2}-\d{2})?$/, "");
+  s = s.replace(/-\d{4}-\d{2}-\d{2}$/, "");
+  s = s.replace(/-\d{8}$/, "");
+  s = s.replace(/-preview$/, "");
+  return s;
+}
+function indexKey(model, provider, region) {
+  return `${provider}|${normalizeForPricing(model)}|${region ?? ""}`;
+}
+var INDEX = /* @__PURE__ */ new Map();
+for (const entry of DATASET.entries) {
+  for (const name of [entry.model, ...entry.aliases ?? []]) {
+    const withRegion = indexKey(name, entry.provider, entry.region);
+    if (!INDEX.has(withRegion)) INDEX.set(withRegion, entry);
+    const bare = indexKey(name, entry.provider);
+    const existing = INDEX.get(bare);
+    if (!existing || existing.region && existing.region !== "global" && entry.region === "global") {
+      INDEX.set(bare, entry);
+    }
+  }
+}
+function findPricing(modelId, provider, opts = {}) {
+  const ids = [modelId, opts.baseModelId].filter((x) => !!x);
+  const providers = [provider, ...opts.fallbackProviders ?? []];
+  for (const p of providers) {
+    for (const id of ids) {
+      const hit = (opts.region ? INDEX.get(indexKey(id, p, opts.region)) : void 0) ?? INDEX.get(indexKey(id, p));
+      if (hit) return { entry: hit, matchedProvider: p, viaFallback: p !== provider };
+    }
+  }
+  return null;
+}
+function tierFor(entry, inputTokens = 0) {
+  const tiers = entry.tiers;
+  if (!tiers?.length) return null;
+  for (const t of tiers) {
+    if (t.maxInputTokens === void 0 || inputTokens <= t.maxInputTokens) return t;
+  }
+  return tiers[tiers.length - 1] ?? null;
+}
+var PER_1M_TO_PER_1K = 1 / 1e3;
+function tokenRatesPer1k(entry, opts = {}) {
+  const tier = tierFor(entry, opts.inputTokens ?? 0);
+  if (!tier) return null;
+  const mult = (opts.batch ? entry.batchMultiplier ?? 1 : 1) * PER_1M_TO_PER_1K;
+  const out = {
+    input: tier.inputPer1m * mult,
+    output: tier.outputPer1m * mult
+  };
+  if (tier.cachedInputPer1m !== void 0) out.cachedInput = tier.cachedInputPer1m * mult;
+  if (tier.cacheWritePer1m !== void 0) out.cacheWrite = tier.cacheWritePer1m * mult;
+  if (tier.reasoningOutputPer1m !== void 0) out.reasoningOutput = tier.reasoningOutputPer1m * mult;
+  return out;
+}
+var LOOPBACK_HOSTS = /* @__PURE__ */ new Set(["localhost", "127.0.0.1", "0.0.0.0", "::1", "[::1]"]);
+function isLoopbackOrPrivateHost(url) {
+  if (!url) return false;
+  try {
+    const host = new URL(url).hostname.toLowerCase().replace(/^\[|\]$/g, "");
+    if (LOOPBACK_HOSTS.has(host) || LOOPBACK_HOSTS.has(`[${host}]`)) return true;
+    if (host.endsWith(".local")) return true;
+    if (/^127\./.test(host)) return true;
+    if (/^10\./.test(host)) return true;
+    if (/^192\.168\./.test(host)) return true;
+    if (/^172\.(1[6-9]|2\d|3[01])\./.test(host)) return true;
+    return false;
+  } catch {
+    return false;
+  }
+}
+function isLocalEndpoint(cfg) {
+  if (typeof cfg.local === "boolean") return cfg.local;
+  if (cfg.type === "ollama") return true;
+  if (cfg.type === "openai-compatible") return isLoopbackOrPrivateHost(cfg.baseUrl);
+  return false;
+}
+var UNKNOWN = { input: 0, output: 0, unknown: true, free: false };
+var FREE = { input: 0, output: 0, unknown: false, free: true };
+function resolvePricing(model, opts = {}) {
+  if (model.isLocal) return FREE;
+  const fallbackProviders = model.provider === "azure" ? ["openai"] : model.provider === "gemini" ? ["vertex"] : [];
+  const match = findPricing(model.id, model.provider, {
+    region: opts.region,
+    baseModelId: model.baseModelId,
+    fallbackProviders
+  });
+  if (!match) return UNKNOWN;
+  const rates = tokenRatesPer1k(match.entry, { inputTokens: opts.inputTokens, batch: opts.batch });
+  if (!rates) return UNKNOWN;
+  return {
+    input: rates.input,
+    output: rates.output,
+    unknown: false,
+    free: false,
+    entry: match.entry,
+    ...match.viaFallback ? { estimatedFromProvider: match.matchedProvider } : {}
+  };
+}
+function withResolvedPricing(model, opts = {}) {
+  const p = resolvePricing(model, opts);
+  return {
+    ...model,
+    inputCostPer1kTokens: p.input,
+    outputCostPer1kTokens: p.output,
+    ...p.unknown ? { pricingUnknown: true } : { pricingUnknown: false }
+  };
+}
+function isPricingUnknown(model) {
+  return model.pricingUnknown === true && !model.isLocal;
+}
+var BLENDED_COST_CEILING = 0.05;
+function blendedCostPer1k(model) {
+  if (isPricingUnknown(model)) return BLENDED_COST_CEILING;
+  return model.inputCostPer1kTokens + model.outputCostPer1kTokens * 2;
+}
+var DISAGREEMENT_THRESHOLD = 0.02;
+function reconcilePrice(model, live) {
+  const baseline = resolvePricing(model);
+  const liveIsZero = !(live.input > 0) && !(live.output > 0);
+  if (liveIsZero && !model.isLocal) {
+    return { input: baseline.input, output: baseline.output, accepted: false };
+  }
+  if (baseline.unknown || !baseline.entry) {
+    return { input: live.input, output: live.output, accepted: true };
+  }
+  const dsBlend = baseline.input + baseline.output * 2;
+  const liveBlend = live.input + live.output * 2;
+  const ratio = dsBlend > 0 ? liveBlend / dsBlend : Number.POSITIVE_INFINITY;
+  const material = dsBlend > 0 && Math.abs(ratio - 1) > DISAGREEMENT_THRESHOLD;
+  return {
+    input: live.input,
+    output: live.output,
+    accepted: true,
+    ...material ? {
+      disagreement: {
+        modelId: model.id,
+        provider: model.provider,
+        dataset: { input: baseline.input, output: baseline.output },
+        live: { input: live.input, output: live.output },
+        ratio: Math.round(ratio * 1e3) / 1e3,
+        datasetAsOf: baseline.entry.asOf,
+        datasetSource: baseline.entry.source
+      }
+    } : {}
+  };
+}
+
 // src/constants.ts
-var CASCADE_VERSION = "0.48.0";
+var CASCADE_VERSION = "0.49.0";
 var CASCADE_CONFIG_DIR = ".cascade";
 var CASCADE_MD_FILE = "CASCADE.md";
 var CASCADE_IGNORE_FILE = ".cascadeignore";
@@ -542,6 +2194,13 @@ var MODELS = {
 for (const _m of Object.values(MODELS)) {
   if (_m.supportsToolUse === void 0) _m.supportsToolUse = !_m.isLocal;
 }
+for (const _m of Object.values(MODELS)) {
+  const priced = resolvePricing(_m);
+  if (priced.unknown || priced.free || priced.estimatedFromProvider) continue;
+  _m.inputCostPer1kTokens = priced.input;
+  _m.outputCostPer1kTokens = priced.output;
+  _m.pricingUnknown = false;
+}
 var T1_MODEL_PRIORITY = [
   "gpt-5.5",
   "gpt-5.4",
@@ -632,21 +2291,49 @@ var PROVIDER_DISPLAY_NAMES = {
 
 // src/utils/cost.ts
 function resolveModelPricing(model) {
-  let input = model.inputCostPer1kTokens;
-  let output = model.outputCostPer1kTokens;
-  if (input === 0 && output === 0 && model.id && !model.isLocal) {
+  if (model.isLocal) return { input: 0, output: 0, unknown: false };
+  if (model.inputCostPer1kTokens > 0 || model.outputCostPer1kTokens > 0) {
+    return {
+      input: model.inputCostPer1kTokens,
+      output: model.outputCostPer1kTokens,
+      unknown: false
+    };
+  }
+  const fromDataset = resolvePricing(model);
+  if (!fromDataset.unknown) {
+    return {
+      input: fromDataset.input,
+      output: fromDataset.output,
+      unknown: false,
+      ...fromDataset.estimatedFromProvider ? { estimatedFromProvider: fromDataset.estimatedFromProvider } : {}
+    };
+  }
+  if (model.id) {
     const known = Object.values(MODELS).find((m) => m.id === model.id && !m.isLocal);
-    if (known) {
-      input = known.inputCostPer1kTokens;
-      output = known.outputCostPer1kTokens;
+    if (known && (known.inputCostPer1kTokens > 0 || known.outputCostPer1kTokens > 0)) {
+      return {
+        input: known.inputCostPer1kTokens,
+        output: known.outputCostPer1kTokens,
+        unknown: false
+      };
     }
   }
-  return { input, output };
+  return { input: 0, output: 0, unknown: true };
 }
 function calculateCost(inputTokens, outputTokens, model) {
   const { input, output } = resolveModelPricing(model);
   return inputTokens / 1e3 * input + outputTokens / 1e3 * output;
 }
+function buildTokenUsage(inputTokens, outputTokens, model) {
+  const { input, output, unknown } = resolveModelPricing(model);
+  return {
+    inputTokens,
+    outputTokens,
+    totalTokens: inputTokens + outputTokens,
+    estimatedCostUsd: inputTokens / 1e3 * input + outputTokens / 1e3 * output,
+    ...unknown ? { costUnknown: true } : {}
+  };
+}
 
 // src/providers/base.ts
 var BaseProvider = class {
@@ -666,12 +2353,7 @@ var BaseProvider = class {
     return calculateCost(inputTokens, outputTokens, this.model);
   }
   makeUsage(inputTokens, outputTokens) {
-    return {
-      inputTokens,
-      outputTokens,
-      totalTokens: inputTokens + outputTokens,
-      estimatedCostUsd: this.estimateCost(inputTokens, outputTokens)
-    };
+    return buildTokenUsage(inputTokens, outputTokens, this.model);
   }
 };
 
@@ -840,7 +2522,7 @@ var AnthropicProvider = class extends BaseProvider {
       return data.data.filter((m) => isChatModel(m.id)).map((m) => {
         const known = Object.values(MODELS).find((km) => km.id === m.id && km.provider === "anthropic");
         if (known) return known;
-        return {
+        return withResolvedPricing({
           id: m.id,
           name: m.display_name || m.id,
           provider: "anthropic",
@@ -851,7 +2533,7 @@ var AnthropicProvider = class extends BaseProvider {
           maxOutputTokens: 8e3,
           supportsStreaming: true,
           isLocal: false
-        };
+        });
       });
     } catch {
       return Object.values(MODELS).filter((m) => m.provider === "anthropic");
@@ -1055,7 +2737,7 @@ var OpenAIProvider = class extends BaseProvider {
       return response.data.filter((m) => isChatModel(m.id)).map((m) => {
         const known = Object.values(MODELS).find((km) => km.id === m.id && km.provider === "openai");
         if (known) return known;
-        return {
+        return withResolvedPricing({
           id: m.id,
           name: m.id,
           provider: "openai",
@@ -1066,7 +2748,7 @@ var OpenAIProvider = class extends BaseProvider {
           maxOutputTokens: 4e3,
           supportsStreaming: true,
           isLocal: false
-        };
+        });
       });
     } catch {
       return Object.values(MODELS).filter((m) => m.provider === "openai");
@@ -1162,6 +2844,10 @@ function azureModelForDeployment(cfg) {
   const baseModelId = cfg.model?.trim() || inferAzureBaseModel(id) || void 0;
   const base = baseModelId ? MODELS[baseModelId] : void 0;
   if (base) {
+    const azurePrice = resolvePricing(
+      { id, provider: "azure", isLocal: false, baseModelId },
+      { region: cfg.region }
+    );
     return {
       ...base,
       id,
@@ -1170,7 +2856,12 @@ function azureModelForDeployment(cfg) {
       provider: "azure",
       baseModelId,
       // real identity for benchmark + live pricing
-      supportsToolUse: base.supportsToolUse ?? true
+      supportsToolUse: base.supportsToolUse ?? true,
+      ...azurePrice.unknown ? {} : {
+        inputCostPer1kTokens: azurePrice.input,
+        outputCostPer1kTokens: azurePrice.output,
+        pricingUnknown: false
+      }
     };
   }
   return {
@@ -1333,7 +3024,7 @@ var GeminiProvider = class extends BaseProvider {
           (km) => km.id === id && km.provider === "gemini"
         );
         if (known) return known;
-        return {
+        return withResolvedPricing({
           id,
           name: m.displayName || id,
           provider: "gemini",
@@ -1344,7 +3035,7 @@ var GeminiProvider = class extends BaseProvider {
           maxOutputTokens: m.outputTokenLimit || 8e3,
           supportsStreaming: true,
           isLocal: false
-        };
+        });
       });
     } catch {
       return Object.values(MODELS).filter((m) => m.provider === "gemini");
@@ -1691,9 +3382,10 @@ var OllamaProvider = class extends BaseProvider {
         return supportedKeywords.some((k) => name.includes(k)) && isChatModel(name);
       });
       const shows = await Promise.all(entries.map((m) => this.showModel(m.name)));
+      const local = isLocalEndpoint({ ...this.config, type: "ollama" });
       return entries.map((m, i) => {
         const show = shows[i];
-        return {
+        return withResolvedPricing({
           id: m.name,
           name: m.name,
           provider: "ollama",
@@ -1703,10 +3395,10 @@ var OllamaProvider = class extends BaseProvider {
           outputCostPer1kTokens: 0,
           maxOutputTokens: 4e3,
           supportsStreaming: true,
-          isLocal: true,
+          isLocal: local,
           supportsToolUse: show?.tools ?? isToolCapable(m.name),
           minSizeB: this.parseSizeB(m.details?.parameter_size)
-        };
+        });
       });
     } catch {
       return [];
@@ -1809,7 +3501,8 @@ var OpenAICompatibleProvider = class extends OpenAIProvider {
     }).filter((x) => typeof x === "string" && x.length > 0);
     if (ids.length === 0) return [this.model];
     const chatIds = ids.filter((id) => isChatModel(id));
-    return (chatIds.length ? chatIds : ids).map((id) => ({
+    const local = isLocalEndpoint(this.config);
+    return (chatIds.length ? chatIds : ids).map((id) => withResolvedPricing({
       id,
       name: id,
       provider: "openai-compatible",
@@ -1819,7 +3512,7 @@ var OpenAICompatibleProvider = class extends OpenAIProvider {
       outputCostPer1kTokens: 0,
       maxOutputTokens: 4e3,
       supportsStreaming: true,
-      isLocal: false
+      isLocal: local
     }));
   }
   async isAvailable() {
@@ -2034,7 +3727,7 @@ var ModelSelector = class {
       else if (this.availableProviders.size === 1) providerStr = Array.from(this.availableProviders)[0];
     }
     if (providerStr && this.availableProviders.has(providerStr)) {
-      const dynamicModel = {
+      const dynamicModel = withResolvedPricing({
         id: actualId,
         name: actualId,
         provider: providerStr,
@@ -2045,7 +3738,7 @@ var ModelSelector = class {
         maxOutputTokens: 8e3,
         supportsStreaming: true,
         isLocal: providerStr === "ollama"
-      };
+      });
       this.addDynamicModel(dynamicModel);
       return dynamicModel;
     }
@@ -2437,6 +4130,8 @@ var LiveDataProvider = class {
   snapshot = null;
   prices = /* @__PURE__ */ new Map();
   capabilities = /* @__PURE__ */ new Map();
+  /** Dataset-vs-live price mismatches, deduped by `provider:modelId`. */
+  disagreements = /* @__PURE__ */ new Map();
   source = "bundled";
   fetchedAt = 0;
   loaded = false;
@@ -2590,14 +4285,45 @@ var LiveDataProvider = class {
   /**
    * Returns a price-corrected copy of each model when live pricing is known,
    * leaving the original untouched (so the shared catalog is never mutated).
+   *
+   * Reconciliation policy (see reconcilePrice in pricing.ts): the bundled
+   * dataset is the authoritative baseline, and a live quote overrides it
+   * because it is fresher by construction — but the disagreement is RECORDED,
+   * not swallowed. A dataset/live mismatch is the best signal available that
+   * the committed prices have gone stale, so it must be visible rather than
+   * silently resolved in either direction. A live quote of $0 for a non-local
+   * model is rejected outright: that is the very failure mode this change
+   * exists to prevent, and a marketplace's free-tier alias must never zero out
+   * the price of the paid model Cascade is actually calling.
    */
   applyLivePricing(models) {
     return models.map((m) => {
       const p = this.getLivePrice(m.baseModelId ?? m.id);
       if (!p) return m;
-      return { ...m, inputCostPer1kTokens: p.input, outputCostPer1kTokens: p.output };
+      const reconciled = reconcilePrice(m, p);
+      if (reconciled.disagreement) this.recordDisagreement(reconciled.disagreement);
+      if (!reconciled.accepted) return m;
+      return {
+        ...m,
+        inputCostPer1kTokens: reconciled.input,
+        outputCostPer1kTokens: reconciled.output,
+        pricingUnknown: false
+      };
     });
   }
+  recordDisagreement(d) {
+    const key = `${d.provider}:${d.modelId}`;
+    if (this.disagreements.has(key)) return;
+    this.disagreements.set(key, d);
+  }
+  /**
+   * Dataset-vs-live price mismatches seen this session, newest-model-first
+   * insertion order. Surfaced by `cascade models --verbose`; a non-empty list
+   * means src/core/router/pricing-data.json is probably due a refresh.
+   */
+  getPriceDisagreements() {
+    return [...this.disagreements.values()];
+  }
   /** Current capability facts for a model id, or null when unknown. */
   getCapability(modelId) {
     return this.capabilities.get(normalizeModelId2(modelId)) ?? null;
@@ -2794,7 +4520,9 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
     tokensByTier: {},
     inputTokensByTier: {},
     outputTokensByTier: {},
-    costByFeature: {}
+    costByFeature: {},
+    untrackedCostCalls: 0,
+    untrackedCostModels: []
   };
   tierModels = /* @__PURE__ */ new Map();
   /**
@@ -3144,7 +4872,7 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
           `Model ${model.id} inference timed out after ${cloudTimeoutMs}ms`
         );
       }
-      const correctedCost = calculateCost(
+      const corrected = buildTokenUsage(
         result.usage.inputTokens,
         result.usage.outputTokens,
         model
@@ -3153,7 +4881,8 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
         ...result,
         usage: {
           ...result.usage,
-          estimatedCostUsd: correctedCost
+          estimatedCostUsd: corrected.estimatedCostUsd,
+          ...corrected.costUnknown ? { costUnknown: true } : { costUnknown: void 0 }
         }
       };
       if (!result || typeof result.content !== "string" || !result.usage) {
@@ -3340,7 +5069,9 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
       tokensByTier: { ...this.stats.tokensByTier },
       inputTokensByTier: { ...this.stats.inputTokensByTier },
       outputTokensByTier: { ...this.stats.outputTokensByTier },
-      costByFeature: { ...this.stats.costByFeature }
+      costByFeature: { ...this.stats.costByFeature },
+      untrackedCostCalls: this.stats.untrackedCostCalls,
+      untrackedCostModels: [...this.stats.untrackedCostModels]
     };
   }
   /**
@@ -3391,7 +5122,9 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
       tokensByTier: {},
       inputTokensByTier: {},
       outputTokensByTier: {},
-      costByFeature: {}
+      costByFeature: {},
+      untrackedCostCalls: 0,
+      untrackedCostModels: []
     };
     this.sessionCostUsd = 0;
     this.budgetState = "ok";
@@ -3576,6 +5309,18 @@ var CascadeRouter = class _CascadeRouter extends EventEmitter {
     if (featureTag) {
       this.stats.costByFeature[featureTag] = (this.stats.costByFeature[featureTag] ?? 0) + usage.estimatedCostUsd;
     }
+    if (usage.costUnknown) {
+      this.stats.untrackedCostCalls += 1;
+      if (!this.stats.untrackedCostModels.includes(model.id)) {
+        this.stats.untrackedCostModels.push(model.id);
+        const capped = this.config?.budget?.maxCostPerRunUsd != null || this.config?.budget?.sessionBudgetUsd != null;
+        this.emit("cost:untracked", {
+          modelId: model.id,
+          provider: model.provider,
+          reason: capped ? `No published price for ${model.provider}:${model.id}. Its spend is not counted toward your cost budget \u2014 the token cap still applies. Add it to the pricing dataset, or set the provider's \`local: true\` if this endpoint really is free.` : `No published price for ${model.provider}:${model.id}. Cost for this model is reported as "not tracked" rather than $0.00.`
+        });
+      }
+    }
     this.runTokens += usage.totalTokens;
     this.runCostUsd += usage.estimatedCostUsd;
     this.updateBudgetState();
@@ -9083,7 +10828,18 @@ var ProviderConfigSchema = z.object({
   apiVersion: z.string().optional(),
   model: z.string().optional(),
   authToken: z.string().optional(),
-  credentialSource: z.string().optional()
+  credentialSource: z.string().optional(),
+  /**
+   * "This endpoint is self-hosted, so inference genuinely costs $0."
+   * Only meaningful for `ollama` and `openai-compatible`. Unset ⇒ inferred:
+   * ollama is local, openai-compatible is local when its baseUrl is
+   * loopback/LAN. Set it explicitly to override — e.g. `false` for an Ollama
+   * instance you rent, `true` for a llama.cpp box on your own LAN reached by
+   * hostname.
+   */
+  local: z.boolean().optional(),
+  /** Region for providers whose price varies by region (Azure: global | us | eu | …). */
+  region: z.string().optional()
 });
 var ModelOverridesSchema = z.object({
   t1: z.string().optional(),
@@ -9669,8 +11425,8 @@ var TaskAnalyzer = class {
   }
   costEfficiency(model, complexity) {
     if (this.tracker) return this.tracker.costEfficiencyScore(model, complexity);
-    const blended = model.inputCostPer1kTokens + model.outputCostPer1kTokens * 2;
-    const normalised = Math.min(1, blended / 0.05);
+    const blended = blendedCostPer1k(model);
+    const normalised = Math.min(1, blended / BLENDED_COST_CEILING);
     const complexityWeight = (6 - complexity) / 5;
     return Math.max(0.1, 1 - normalised * complexityWeight);
   }
@@ -9805,8 +11561,8 @@ var ModelPerformanceTracker = class {
    * normalised over $0.05 blended as the "expensive" ceiling.
    */
   costEfficiencyScore(model, complexity) {
-    const blended = model.inputCostPer1kTokens + model.outputCostPer1kTokens * 2;
-    const normalised = Math.min(1, blended / 0.05);
+    const blended = blendedCostPer1k(model);
+    const normalised = Math.min(1, blended / BLENDED_COST_CEILING);
     const complexityWeight = (6 - complexity) / 5;
     return Math.max(0.1, 1 - normalised * complexityWeight);
   }
@@ -12649,7 +14405,11 @@ _(incomplete: ${reasons.join("; ")})_` : partial;
         inputTokens: 0,
         outputTokens: 0,
         totalTokens: stats.totalTokens,
-        estimatedCostUsd: stats.totalCostUsd
+        estimatedCostUsd: stats.totalCostUsd,
+        // At least one call ran on a model with no known price, so its spend is
+        // missing from the total above. Say so rather than letting the caller
+        // present an undercount as the final figure.
+        ...stats.untrackedCostCalls > 0 ? { costUnknown: true } : {}
       },
       t2Results,
       durationMs,
@@ -12719,7 +14479,8 @@ _(incomplete: ${reasons.join("; ")})_` : partial;
             inputTokens: 0,
             outputTokens: 0,
             totalTokens: stats2.totalTokens,
-            estimatedCostUsd: stats2.totalCostUsd
+            estimatedCostUsd: stats2.totalCostUsd,
+            ...stats2.untrackedCostCalls > 0 ? { costUnknown: true } : {}
           },
           t2Results: [],
           durationMs: Date.now() - startMs,
@@ -12743,7 +14504,8 @@ _(incomplete: ${reasons.join("; ")})_` : partial;
         inputTokens: result.usage?.inputTokens ?? 0,
         outputTokens: result.usage?.outputTokens ?? 0,
         totalTokens: stats.totalTokens,
-        estimatedCostUsd: stats.totalCostUsd
+        estimatedCostUsd: stats.totalCostUsd,
+        ...stats.untrackedCostCalls > 0 ? { costUnknown: true } : {}
       },
       t2Results: [],
       durationMs: Date.now() - startMs,
@@ -14527,7 +16289,10 @@ var DashboardServer = class {
         this.socket.broadcast("cost:update", {
           sessionId,
           totalTokens: result.usage.totalTokens,
-          totalCostUsd: result.usage.estimatedCostUsd
+          totalCostUsd: result.usage.estimatedCostUsd,
+          // True when a model with no known price was used: the total above is
+          // an undercount, and the UI must not present it as the full spend.
+          costUnknown: result.usage.costUnknown === true
         });
         this.throttledBroadcast("workspace");
       } catch (err) {
@@ -14840,7 +16605,8 @@ var DashboardServer = class {
       this.socket.broadcast("cost:update", {
         sessionId,
         totalTokens: result.usage.totalTokens,
-        totalCostUsd: result.usage.estimatedCostUsd
+        totalCostUsd: result.usage.estimatedCostUsd,
+        costUnknown: result.usage.costUnknown === true
       });
       this.throttledBroadcast("workspace");
     } catch (err) {
@@ -15382,7 +17148,8 @@ ${prompt}`;
           this.socket.broadcast("cost:update", {
             sessionId,
             totalTokens: result.usage.totalTokens,
-            totalCostUsd: result.usage.estimatedCostUsd
+            totalCostUsd: result.usage.estimatedCostUsd,
+            costUnknown: result.usage.costUnknown === true
           });
           this.socket.broadcastToRoom(`session:${sessionId}`, "session:complete", { sessionId, result });
           this.throttledBroadcast("workspace");
@@ -15938,12 +17705,19 @@ var CloudClient = class _CloudClient {
       );
       await openUrl(authorizeUrl);
       const code = await listener.waitForCode;
-      const res = await fetch(this.url("/api/native/token"), {
-        method: "POST",
-        headers: { "Content-Type": "application/json" },
-        body: JSON.stringify({ code, code_verifier: verifier })
-      });
-      if (!res.ok) throw new Error("Sign-in could not be completed. Please try again.");
+      let res;
+      try {
+        res = await fetch(this.url("/api/native/token"), {
+          method: "POST",
+          headers: { "Content-Type": "application/json" },
+          body: JSON.stringify({ code, code_verifier: verifier })
+        });
+      } catch (err) {
+        throw new Error(
+          `Signed in, but could not reach Cascade at ${this.serverUrl} to finish. Check your connection and that the server URL is correct. (${err instanceof Error ? err.message : String(err)})`
+        );
+      }
+      if (!res.ok) throw new Error(await loginFailureMessage(res, this.serverUrl));
       const tokens = await res.json();
       return await this.finishLogin(tokens);
     } finally {
@@ -16117,6 +17891,22 @@ function sleep(ms, signal) {
     }, { once: true });
   });
 }
+async function loginFailureMessage(res, serverUrl) {
+  const body = await res.text().catch(() => "");
+  let apiError = "";
+  try {
+    apiError = String(JSON.parse(body).error ?? "");
+  } catch {
+  }
+  if (apiError === "invalid_grant") {
+    return "Sign-in expired before it completed \u2014 the code is single-use and short-lived, and restarting the server during sign-in also invalidates it. Please try again.";
+  }
+  if (apiError) return `Sign-in failed: ${apiError} (HTTP ${res.status}).`;
+  if (res.status === 404 || /<html/i.test(body)) {
+    return `Signed in, but ${serverUrl} did not answer the Cascade API (HTTP ${res.status}). If this host was recently pointed at a new server, DNS may still be propagating.`;
+  }
+  return `Sign-in could not be completed (HTTP ${res.status}). Please try again.`;
+}
 function base64url(buf) {
   return buf.toString("base64").replace(/\+/g, "-").replace(/\//g, "_").replace(/=+$/, "");
 }