@azure-id/orc 1.9.1 → 2.0.2

package/bin/pricing.json

@@ -1,200 +1,207 @@
-{
-  "_readme": [
-    "Dated price table for `orc budget`. Point `orc config set budget_price_table <path>`",
-    "at your own copy to override it. Rates are USD per MILLION tokens.",
-    "",
-    "Four rates per model, never one: cache reads are usually the LARGEST token",
-    "count in a run and the CHEAPEST per token, so a blended number hides",
-    "whichever component is about to bite you. cache_write/cache_read are the",
-    "multipliers this vendor publishes against `input` (1.25x / 0.1x) written out",
-    "as absolute rates so nothing has to remember the ratio.",
-    "",
-    "`as_of` is load-bearing: a table older than 90 days makes every dollar",
-    "figure print a staleness warning. Bump the date only when you have checked",
-    "the published rates.",
-    "",
-    "`providers` (v0.50.0) is the same four-rate shape for a NON-CLAUDE endpoint",
-    "reached through `orc extra`. Every `models` map SHIPS EMPTY, on purpose, and",
-    "that is not an oversight: between this package's cutoff and any release the",
-    "lineup moves (DeepSeek V3.2 -> V4, Kimi K2 -> K2.7, GLM 4.6 -> 5.2), several",
-    "of these vendors price by PEAK/OFF-PEAK window or by tier, one is",
-    "subscription-only, and one is a passthrough with a surcharge. A shipped",
-    "figure that is wrong by 2x is worse than no figure, because a wrong figure",
-    "gets believed. So ORC ships the MECHANISM and no dollar figure without a",
-    "rate you put there yourself.",
-    "",
-    "Fill one in with `orc extra rates` — it lists every provider/model pair your",
-    "own traces have actually used, prints the JSON to paste, and says which ones",
-    "it has no rate for. A pair with no rate reports `usd: null` forever, never a",
-    "guess and never a family fallback."
-  ],
-  "as_of": "2026-08-01",
-  "currency": "USD",
-  "unit": "per_million_tokens",
-  "server_tools": {
-    "web_search_requests": 0.01,
-    "web_fetch_requests": 0.0
-  },
-  "models": {
-    "claude-opus-5": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-opus-4-8": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-opus-4-7": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-fable-5": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "claude-sonnet-5": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "claude-sonnet-4-6": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "claude-haiku-4-5": {
-      "input": 1,
-      "cache_write": 1.25,
-      "cache_read": 0.1,
-      "output": 5
-    }
-  },
-  "families": {
-    "opus": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "fable": {
-      "input": 15,
-      "cache_write": 18.75,
-      "cache_read": 1.5,
-      "output": 75
-    },
-    "sonnet": {
-      "input": 3,
-      "cache_write": 3.75,
-      "cache_read": 0.3,
-      "output": 15
-    },
-    "haiku": {
-      "input": 1,
-      "cache_write": 1.25,
-      "cache_read": 0.1,
-      "output": 5
-    }
-  },
-  "_plans_readme": [
-    "Quota view. A Pro/Max user does not pay per token — they burn a 5-hour",
-    "session window and a weekly ceiling. These are WEIGHTED-token capacities",
-    "(cache reads counted at their 0.1x weight), because that is the unit the",
-    "forecast can actually produce. They are approximations of a published",
-    "allowance, not a metered balance: the percentages are a magnitude, and the",
-    "renderer says so."
-  ],
-  "plans": {
-    "pro": {
-      "label": "Pro",
-      "window_weighted_tokens": 900000,
-      "weekly_weighted_tokens": 9000000
-    },
-    "max5": {
-      "label": "Max 5x",
-      "window_weighted_tokens": 4500000,
-      "weekly_weighted_tokens": 45000000
-    },
-    "max20": {
-      "label": "Max 20x",
-      "window_weighted_tokens": 18000000,
-      "weekly_weighted_tokens": 180000000
-    }
-  },
-  "context_windows": {
-    "claude-opus-5": 200000,
-    "claude-opus-4-8": 200000,
-    "claude-opus-4-7": 200000,
-    "claude-fable-5": 200000,
-    "claude-sonnet-5": 200000,
-    "claude-sonnet-4-6": 200000,
-    "claude-haiku-4-5": 200000
-  },
-  "providers": {
-    "_note": "Rates are USD per MILLION tokens, four per model, never one. `models` maps a provider's own model id EXACTLY as the endpoint reports it.",
-    "deepseek": {
-      "label": "DeepSeek",
-      "caveat": "Prices differ by PEAK and OFF-PEAK window, so one number cannot be right for both. Record the window you actually run in.",
-      "models": {}
-    },
-    "zai": {
-      "label": "Z.ai / GLM",
-      "caveat": "The current flagship is sold inside a Coding Plan subscription rather than per token, which a per-token table cannot express. Leave it empty and read your invoice.",
-      "models": {}
-    },
-    "moonshot": {
-      "label": "Moonshot / Kimi",
-      "caveat": null,
-      "models": {}
-    },
-    "minimax": {
-      "label": "MiniMax",
-      "caveat": null,
-      "models": {}
-    },
-    "qwen": {
-      "label": "Alibaba / Qwen",
-      "caveat": "Tiered by request size, so a single rate is an average rather than a price.",
-      "models": {}
-    },
-    "mimo": {
-      "label": "Xiaomi / MiMo",
-      "caveat": null,
-      "models": {}
-    },
-    "stepfun": {
-      "label": "StepFun",
-      "caveat": null,
-      "models": {}
-    },
-    "siliconflow": {
-      "label": "SiliconFlow",
-      "caveat": null,
-      "models": {}
-    },
-    "openrouter": {
-      "label": "OpenRouter",
-      "caveat": "A PASSTHROUGH: the rate is the upstream provider's, plus a surcharge on credit purchase. It also reroutes between providers by default while preserving the model id, so the model id alone does not identify who you paid.",
-      "models": {}
-    },
-    "ollama": {
-      "label": "Ollama (local)",
-      "caveat": "Local inference has no per-token price. Zero here would be true of the API bill and false of the electricity and the hardware, so it is left empty rather than asserted as free.",
-      "models": {}
-    },
-    "custom": {
-      "label": "Custom endpoint",
-      "caveat": null,
-      "models": {}
-    }
-  }
-}
+{
+  "_readme": [
+    "Dated price table for `orc budget`. Point `orc config set budget_price_table <path>`",
+    "at your own copy to override it. Rates are USD per MILLION tokens.",
+    "",
+    "Four rates per model, never one: cache reads are usually the LARGEST token",
+    "count in a run and the CHEAPEST per token, so a blended number hides",
+    "whichever component is about to bite you. cache_write/cache_read are the",
+    "multipliers this vendor publishes against `input` (1.25x / 0.1x) written out",
+    "as absolute rates so nothing has to remember the ratio.",
+    "",
+    "`as_of` is load-bearing: a table older than 90 days makes every dollar",
+    "figure print a staleness warning. Bump the date only when you have checked",
+    "the published rates.",
+    "",
+    "`providers` (v0.50.0) is the same four-rate shape for a NON-CLAUDE endpoint",
+    "reached through `orc extra`. Every `models` map SHIPS EMPTY, on purpose, and",
+    "that is not an oversight: between this package's cutoff and any release the",
+    "lineup moves (DeepSeek V3.2 -> V4, Kimi K2 -> K2.7, GLM 4.6 -> 5.2), several",
+    "of these vendors price by PEAK/OFF-PEAK window or by tier, one is",
+    "subscription-only, and one is a passthrough with a surcharge. A shipped",
+    "figure that is wrong by 2x is worse than no figure, because a wrong figure",
+    "gets believed. So ORC ships the MECHANISM and no dollar figure without a",
+    "rate you put there yourself.",
+    "",
+    "Fill one in with `orc extra rates` — it lists every provider/model pair your",
+    "own traces have actually used, prints the JSON to paste, and says which ones",
+    "it has no rate for. A pair with no rate reports `usd: null` forever, never a",
+    "guess and never a family fallback."
+  ],
+  "as_of": "2026-08-01",
+  "currency": "USD",
+  "unit": "per_million_tokens",
+  "server_tools": {
+    "web_search_requests": 0.01,
+    "web_fetch_requests": 0.0
+  },
+  "models": {
+    "claude-opus-5-5": {
+      "input": 4,
+      "cache_write": 5,
+      "cache_read": 0.2,
+      "output": 20
+    },
+    "claude-opus-5": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-opus-4-8": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-opus-4-7": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-fable-5": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "claude-sonnet-5": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "claude-sonnet-4-6": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "claude-haiku-4-5": {
+      "input": 1,
+      "cache_write": 1.25,
+      "cache_read": 0.1,
+      "output": 5
+    }
+  },
+  "families": {
+    "opus": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "fable": {
+      "input": 15,
+      "cache_write": 18.75,
+      "cache_read": 1.5,
+      "output": 75
+    },
+    "sonnet": {
+      "input": 3,
+      "cache_write": 3.75,
+      "cache_read": 0.3,
+      "output": 15
+    },
+    "haiku": {
+      "input": 1,
+      "cache_write": 1.25,
+      "cache_read": 0.1,
+      "output": 5
+    }
+  },
+  "_plans_readme": [
+    "Quota view. A Pro/Max user does not pay per token — they burn a 5-hour",
+    "session window and a weekly ceiling. These are WEIGHTED-token capacities",
+    "(cache reads counted at their 0.1x weight), because that is the unit the",
+    "forecast can actually produce. They are approximations of a published",
+    "allowance, not a metered balance: the percentages are a magnitude, and the",
+    "renderer says so."
+  ],
+  "plans": {
+    "pro": {
+      "label": "Pro",
+      "window_weighted_tokens": 900000,
+      "weekly_weighted_tokens": 9000000
+    },
+    "max5": {
+      "label": "Max 5x",
+      "window_weighted_tokens": 4500000,
+      "weekly_weighted_tokens": 45000000
+    },
+    "max20": {
+      "label": "Max 20x",
+      "window_weighted_tokens": 18000000,
+      "weekly_weighted_tokens": 180000000
+    }
+  },
+  "context_windows": {
+    "claude-opus-5-5": 200000,
+    "claude-opus-5": 200000,
+    "claude-opus-4-8": 200000,
+    "claude-opus-4-7": 200000,
+    "claude-fable-5": 200000,
+    "claude-sonnet-5": 200000,
+    "claude-sonnet-4-6": 200000,
+    "claude-haiku-4-5": 200000
+  },
+  "providers": {
+    "_note": "Rates are USD per MILLION tokens, four per model, never one. `models` maps a provider's own model id EXACTLY as the endpoint reports it.",
+    "deepseek": {
+      "label": "DeepSeek",
+      "caveat": "Prices differ by PEAK and OFF-PEAK window, so one number cannot be right for both. Record the window you actually run in.",
+      "models": {}
+    },
+    "zai": {
+      "label": "Z.ai / GLM",
+      "caveat": "The current flagship is sold inside a Coding Plan subscription rather than per token, which a per-token table cannot express. Leave it empty and read your invoice.",
+      "models": {}
+    },
+    "moonshot": {
+      "label": "Moonshot / Kimi",
+      "caveat": null,
+      "models": {}
+    },
+    "minimax": {
+      "label": "MiniMax",
+      "caveat": null,
+      "models": {}
+    },
+    "qwen": {
+      "label": "Alibaba / Qwen",
+      "caveat": "Tiered by request size, so a single rate is an average rather than a price.",
+      "models": {}
+    },
+    "mimo": {
+      "label": "Xiaomi / MiMo",
+      "caveat": null,
+      "models": {}
+    },
+    "stepfun": {
+      "label": "StepFun",
+      "caveat": null,
+      "models": {}
+    },
+    "siliconflow": {
+      "label": "SiliconFlow",
+      "caveat": null,
+      "models": {}
+    },
+    "openrouter": {
+      "label": "OpenRouter",
+      "caveat": "A PASSTHROUGH: the rate is the upstream provider's, plus a surcharge on credit purchase. It also reroutes between providers by default while preserving the model id, so the model id alone does not identify who you paid.",
+      "models": {}
+    },
+    "ollama": {
+      "label": "Ollama (local)",
+      "caveat": "Local inference has no per-token price. Zero here would be true of the API bill and false of the electricity and the hardware, so it is left empty rather than asserted as free.",
+      "models": {}
+    },
+    "custom": {
+      "label": "Custom endpoint",
+      "caveat": null,
+      "models": {}
+    }
+  }
+}