@gajae-code/ai 0.7.8 → 0.7.10

package/CHANGELOG.md

@@ -2,6 +2,12 @@
 
 ## [Unreleased]
 
+## [0.7.9] - 2026-07-01
+
+### Fixed
+
+- Mapped DeepSeek-style `prompt_cache_hit_tokens` and `prompt_cache_miss_tokens` usage fields into OpenAI-compatible prompt-cache accounting (#1329).
+
 ## [0.7.5] - 2026-06-27
 
 ### Fixed

package/dist/types/provider-models/openai-compat.d.ts

@@ -58,6 +58,11 @@ export interface DeepSeekModelManagerConfig {
     baseUrl?: string;
 }
 export declare function deepseekModelManagerOptions(config?: DeepSeekModelManagerConfig): ModelManagerOptions<"openai-completions">;
+export interface DeepInfraModelManagerConfig {
+    apiKey?: string;
+    baseUrl?: string;
+}
+export declare function deepinfraModelManagerOptions(config?: DeepInfraModelManagerConfig): ModelManagerOptions<"openai-completions">;
 export interface FireworksModelManagerConfig {
     apiKey?: string;
     baseUrl?: string;

package/dist/types/providers/anthropic.d.ts

@@ -134,6 +134,7 @@ export type AnthropicClientOptionsArgs = {
     onSseEvent?: AnthropicOptions["onSseEvent"];
     fetch?: FetchImpl;
     requestMaxRetries?: number;
+    maxRetryDelayMs?: number;
 };
 export type AnthropicClientOptionsResult = {
     isOAuthToken: boolean;

package/dist/types/providers/google-gemini-headers.d.ts

@@ -3,6 +3,9 @@
  * Uses the same format as the official Gemini CLI (v0.35+):
  * GeminiCLI/VERSION/MODEL (PLATFORM; ARCH; SURFACE)
  */
+export declare const GEMINI_CLI_VERSION_ENV = "GJC_AI_GEMINI_CLI_VERSION";
+export declare const LEGACY_GEMINI_CLI_VERSION_ENV = "PI_AI_GEMINI_CLI_VERSION";
+export declare const DEFAULT_GEMINI_CLI_VERSION = "0.49.0";
 export declare function getGeminiCliUserAgent(modelId?: string): string;
 export declare const getGeminiCliHeaders: (modelId?: string) => {
     "User-Agent": string;

package/dist/types/types.d.ts

@@ -48,7 +48,7 @@ export interface ThinkingConfig {
     /** Provider-specific transport used to encode the selected effort. */
     mode: ThinkingControlMode;
 }
-export type KnownProvider = "alibaba-coding-plan" | "amazon-bedrock" | "azure-openai" | "anthropic" | "google" | "google-gemini-cli" | "google-antigravity" | "google-vertex" | "openai" | "openai-codex" | "kimi-code" | "minimax-code" | "minimax-code-cn" | "github-copilot" | "fireworks" | "firepass" | "fugu" | "gitlab-duo" | "cursor" | "deepseek" | "xai" | "groq" | "cerebras" | "openrouter" | "kilo" | "vercel-ai-gateway" | "zai" | "glm-zcode" | "mistral" | "minimax" | "opencode-go" | "opencode-zen" | "synthetic" | "cloudflare-ai-gateway" | "huggingface" | "litellm" | "moonshot" | "nvidia" | "nanogpt" | "ollama" | "ollama-cloud" | "qianfan" | "qwen-portal" | "together" | "venice" | "vllm" | "xiaomi" | "xiaomi-token-plan-sgp" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "zenmux" | "lm-studio";
+export type KnownProvider = "alibaba-coding-plan" | "amazon-bedrock" | "azure-openai" | "anthropic" | "google" | "google-gemini-cli" | "google-antigravity" | "google-vertex" | "openai" | "openai-codex" | "kimi-code" | "minimax-code" | "minimax-code-cn" | "github-copilot" | "fireworks" | "firepass" | "fugu" | "gitlab-duo" | "cursor" | "deepseek" | "deepinfra" | "xai" | "groq" | "cerebras" | "openrouter" | "kilo" | "vercel-ai-gateway" | "zai" | "glm-zcode" | "mistral" | "minimax" | "opencode-go" | "opencode-zen" | "synthetic" | "cloudflare-ai-gateway" | "huggingface" | "litellm" | "moonshot" | "nvidia" | "nanogpt" | "ollama" | "ollama-cloud" | "qianfan" | "qwen-portal" | "together" | "venice" | "vllm" | "xiaomi" | "xiaomi-token-plan-sgp" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "zenmux" | "lm-studio";
 export type Provider = KnownProvider | string;
 import type { Effort } from "./model-thinking";
 /** Token budgets for each thinking level (token-based providers only) */
@@ -84,7 +84,7 @@ export type CacheRetention = "none" | "short" | "long";
  *
  * The unscoped values (`"auto"`, `"default"`, `"flex"`, `"scale"`,
  * `"priority"`) are passed through to providers that understand them
- * (OpenAI's `service_tier` field directly; Anthropic translates
+ * (OpenAI and DeepInfra's `service_tier` field directly; Anthropic translates
  * `"priority"` into `speed: "fast"` on supported Opus models).
  *
  * The scoped values target a specific provider family and behave as the
@@ -105,10 +105,9 @@ export type ResolvedServiceTier = Exclude<ServiceTier, "openai-only" | "claude-o
  */
 export declare function resolveServiceTier(serviceTier: ServiceTier | null | undefined, provider: Provider | undefined): ResolvedServiceTier | undefined;
 /**
- * True when the (possibly scoped) tier should be sent as OpenAI's
- * `service_tier` request field for the given provider. Non-OpenAI
- * providers, unsupported tiers (`"auto"`, `"default"`), and scope
- * mismatches all return false.
+ * True when the (possibly scoped) tier should be sent as an OpenAI-compatible
+ * `service_tier` request field for providers that support it. Unsupported tiers
+ * (`"auto"`, `"default"`) and scope mismatches all return false.
  */
 export declare function shouldSendServiceTier(serviceTier: ServiceTier | null | undefined, provider: Provider | undefined): boolean;
 /**

package/dist/types/utils/oauth/deepinfra.d.ts

@@ -0,0 +1 @@
+export declare const loginDeepInfra: (options: import("./types").OAuthController) => Promise<string>;

package/dist/types/utils/oauth/types.d.ts

@@ -7,7 +7,7 @@ export type OAuthCredentials = {
     email?: string;
     accountId?: string;
 };
-export type OAuthProvider = "alibaba-coding-plan" | "anthropic" | "cerebras" | "cloudflare-ai-gateway" | "cursor" | "deepseek" | "fireworks" | "firepass" | "fugu" | "github-copilot" | "google-gemini-cli" | "google-antigravity" | "gitlab-duo" | "huggingface" | "kimi-code" | "kilo" | "kagi" | "litellm" | "lm-studio" | "minimax-code" | "minimax-code-cn" | "moonshot" | "nvidia" | "nanogpt" | "ollama" | "ollama-cloud" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "synthetic" | "tavily" | "together" | "venice" | "vercel-ai-gateway" | "vllm" | "xai" | "glm-zcode" | "xiaomi" | "xiaomi-token-plan-sgp" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "zenmux" | "zai";
+export type OAuthProvider = "alibaba-coding-plan" | "anthropic" | "cerebras" | "cloudflare-ai-gateway" | "cursor" | "deepseek" | "deepinfra" | "fireworks" | "firepass" | "fugu" | "github-copilot" | "google-gemini-cli" | "google-antigravity" | "gitlab-duo" | "huggingface" | "kimi-code" | "kilo" | "kagi" | "litellm" | "lm-studio" | "minimax-code" | "minimax-code-cn" | "moonshot" | "nvidia" | "nanogpt" | "ollama" | "ollama-cloud" | "openai-codex" | "openai-codex-device" | "opencode-go" | "opencode-zen" | "parallel" | "perplexity" | "qianfan" | "qwen-portal" | "synthetic" | "tavily" | "together" | "venice" | "vercel-ai-gateway" | "vllm" | "xai" | "glm-zcode" | "xiaomi" | "xiaomi-token-plan-sgp" | "xiaomi-token-plan-ams" | "xiaomi-token-plan-cn" | "zenmux" | "zai";
 export type OAuthProviderId = OAuthProvider | (string & {});
 export type OAuthPrompt = {
     message: string;

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@gajae-code/ai",
-	"version": "0.7.8",
+	"version": "0.7.10",
 	"description": "Unified LLM API with automatic model discovery and provider configuration",
 	"homepage": "https://gaebal-gajae.dev",
 	"author": "Yeachan-Heo",
@@ -43,7 +43,7 @@
 	"dependencies": {
 		"@anthropic-ai/sdk": "^0.94.0",
 		"@bufbuild/protobuf": "^2.12.0",
-		"@gajae-code/utils": "0.7.8",
+		"@gajae-code/utils": "0.7.10",
 		"openai": "^6.36.0",
 		"partial-json": "^0.1.7",
 		"zod": "4.4.3"

package/src/auth-storage.ts

@@ -32,6 +32,7 @@ import { kimiUsageProvider } from "./usage/kimi";
 import { codexRankingStrategy, openaiCodexUsageProvider } from "./usage/openai-codex";
 import { zaiUsageProvider } from "./usage/zai";
 import { getOAuthApiKey, getOAuthProvider, refreshOAuthToken, resolveOAuthStorageProvider } from "./utils/oauth";
+import { loginDeepInfra } from "./utils/oauth/deepinfra";
 import { loginDeepSeek } from "./utils/oauth/deepseek";
 import { loginOpenAICodexDevice } from "./utils/oauth/openai-codex";
 import type { OAuthController, OAuthCredentials, OAuthProvider, OAuthProviderId } from "./utils/oauth/types";
@@ -1593,6 +1594,11 @@ export class AuthStorage {
 				await saveApiKeyCredential(apiKey);
 				return;
 			}
+			case "deepinfra": {
+				const apiKey = await loginDeepInfra(ctrl);
+				await saveApiKeyCredential(apiKey);
+				return;
+			}
 			case "xai": {
 				const { loginXai } = await import("./utils/oauth/xai");
 				credentials = await loginXai({

package/src/cli.ts

@@ -111,6 +111,7 @@ Providers:
   zai               Z.AI (GLM Coding Plan)
   glm-zcode         GLM ZCode OAuth (unofficial, opt-in; at your own risk)
   deepseek          DeepSeek
+  deepinfra         DeepInfra
   xai               xAI
   nanogpt           NanoGPT
   minimax-code      MiniMax Coding Plan (International)

package/src/models.json

@@ -3583,6 +3583,31 @@
 				"minLevel": "minimal",
 				"maxLevel": "high"
 			}
+		},
+		"claude-sonnet-5": {
+			"id": "claude-sonnet-5",
+			"name": "Anthropic Sonnet 5",
+			"api": "anthropic-messages",
+			"provider": "anthropic",
+			"baseUrl": "https://api.anthropic.com",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 3,
+				"output": 15,
+				"cacheRead": 0.3,
+				"cacheWrite": 3.75
+			},
+			"contextWindow": 1000000,
+			"maxTokens": 64000,
+			"thinking": {
+				"mode": "anthropic-adaptive",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
 		}
 	},
 	"azure-openai": {
@@ -7722,6 +7747,602 @@
 			}
 		}
 	},
+	"deepinfra": {
+		"deepseek-ai/DeepSeek-R1-0528": {
+			"id": "deepseek-ai/DeepSeek-R1-0528",
+			"name": "DeepSeek-R1-0528",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.5,
+				"output": 2.15,
+				"cacheRead": 0.35,
+				"cacheWrite": 0
+			},
+			"contextWindow": 163840,
+			"maxTokens": 64000,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"deepseek-ai/DeepSeek-V3.2": {
+			"id": "deepseek-ai/DeepSeek-V3.2",
+			"name": "DeepSeek-V3.2",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.26,
+				"output": 0.38,
+				"cacheRead": 0.13,
+				"cacheWrite": 0
+			},
+			"contextWindow": 163840,
+			"maxTokens": 64000,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"deepseek-ai/DeepSeek-V4-Flash": {
+			"id": "deepseek-ai/DeepSeek-V4-Flash",
+			"name": "DeepSeek V4 Flash",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.1,
+				"output": 0.2,
+				"cacheRead": 0.02,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"deepseek-ai/DeepSeek-V4-Pro": {
+			"id": "deepseek-ai/DeepSeek-V4-Pro",
+			"name": "DeepSeek V4 Pro",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 1.3,
+				"output": 2.6,
+				"cacheRead": 0.1,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"google/gemma-4-26B-A4B-it": {
+			"id": "google/gemma-4-26B-A4B-it",
+			"name": "Gemma 4 26B A4B IT",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.07,
+				"output": 0.34,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 32768,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"google/gemma-4-31B-it": {
+			"id": "google/gemma-4-31B-it",
+			"name": "Gemma 4 31B IT",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.13,
+				"output": 0.38,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 32768,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"meta-llama/Llama-3.3-70B-Instruct-Turbo": {
+			"id": "meta-llama/Llama-3.3-70B-Instruct-Turbo",
+			"name": "Llama 3.3 70B Turbo",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.1,
+				"output": 0.32,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 131072,
+			"maxTokens": 16384
+		},
+		"meta-llama/Llama-4-Scout-17B-16E-Instruct": {
+			"id": "meta-llama/Llama-4-Scout-17B-16E-Instruct",
+			"name": "Llama 4 Scout 17B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": false,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.1,
+				"output": 0.3,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 327680,
+			"maxTokens": 16384
+		},
+		"MiniMaxAI/MiniMax-M2.5": {
+			"id": "MiniMaxAI/MiniMax-M2.5",
+			"name": "MiniMax M2.5",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.15,
+				"output": 1.15,
+				"cacheRead": 0.03,
+				"cacheWrite": 0.375
+			},
+			"contextWindow": 196608,
+			"maxTokens": 131072,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"moonshotai/Kimi-K2.5": {
+			"id": "moonshotai/Kimi-K2.5",
+			"name": "Kimi K2.5",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.45,
+				"output": 2.25,
+				"cacheRead": 0.07,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 32768,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"moonshotai/Kimi-K2.6": {
+			"id": "moonshotai/Kimi-K2.6",
+			"name": "Kimi K2.6",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.75,
+				"output": 3.5,
+				"cacheRead": 0.15,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"openai/gpt-oss-120b": {
+			"id": "openai/gpt-oss-120b",
+			"name": "GPT OSS 120B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.039,
+				"output": 0.19,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 131072,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"openai/gpt-oss-20b": {
+			"id": "openai/gpt-oss-20b",
+			"name": "GPT OSS 20B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.03,
+				"output": 0.14,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 131072,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo": {
+			"id": "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo",
+			"name": "Qwen3 Coder 480B A35B Instruct Turbo",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": false,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.3,
+				"output": 1,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 66536
+		},
+		"Qwen/Qwen3.5-35B-A3B": {
+			"id": "Qwen/Qwen3.5-35B-A3B",
+			"name": "Qwen 3.5 35B A3B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.14,
+				"output": 1,
+				"cacheRead": 0.05,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 81920,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
+		},
+		"Qwen/Qwen3.5-397B-A17B": {
+			"id": "Qwen/Qwen3.5-397B-A17B",
+			"name": "Qwen 3.5 397B A17B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.45,
+				"output": 3,
+				"cacheRead": 0.22,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 81920,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
+		},
+		"Qwen/Qwen3.6-35B-A3B": {
+			"id": "Qwen/Qwen3.6-35B-A3B",
+			"name": "Qwen3.6 35B A3B",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.15,
+				"output": 0.95,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 81920,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "high"
+			}
+		},
+		"XiaomiMiMo/MiMo-V2.5": {
+			"id": "XiaomiMiMo/MiMo-V2.5",
+			"name": "MiMo-V2.5",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text",
+				"image"
+			],
+			"cost": {
+				"input": 0.4,
+				"output": 2,
+				"cacheRead": 0.08,
+				"cacheWrite": 0
+			},
+			"contextWindow": 262144,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"XiaomiMiMo/MiMo-V2.5-Pro": {
+			"id": "XiaomiMiMo/MiMo-V2.5-Pro",
+			"name": "MiMo-V2.5-Pro",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 1,
+				"output": 3,
+				"cacheRead": 0.2,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-4.6": {
+			"id": "zai-org/GLM-4.6",
+			"name": "GLM-4.6",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.43,
+				"output": 1.74,
+				"cacheRead": 0.08,
+				"cacheWrite": 0
+			},
+			"contextWindow": 202752,
+			"maxTokens": 131072,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-4.7": {
+			"id": "zai-org/GLM-4.7",
+			"name": "GLM-4.7",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.4,
+				"output": 1.75,
+				"cacheRead": 0.08,
+				"cacheWrite": 0
+			},
+			"contextWindow": 202752,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-4.7-Flash": {
+			"id": "zai-org/GLM-4.7-Flash",
+			"name": "GLM-4.7-Flash",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.06,
+				"output": 0.4,
+				"cacheRead": 0,
+				"cacheWrite": 0
+			},
+			"contextWindow": 202752,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-5": {
+			"id": "zai-org/GLM-5",
+			"name": "GLM-5",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.6,
+				"output": 2.08,
+				"cacheRead": 0.12,
+				"cacheWrite": 0
+			},
+			"contextWindow": 202752,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-5.1": {
+			"id": "zai-org/GLM-5.1",
+			"name": "GLM-5.1",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 1.05,
+				"output": 3.5,
+				"cacheRead": 0.205,
+				"cacheWrite": 0
+			},
+			"contextWindow": 202752,
+			"maxTokens": 16384,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		},
+		"zai-org/GLM-5.2": {
+			"id": "zai-org/GLM-5.2",
+			"name": "GLM-5.2",
+			"api": "openai-completions",
+			"provider": "deepinfra",
+			"baseUrl": "https://api.deepinfra.com/v1/openai",
+			"reasoning": true,
+			"input": [
+				"text"
+			],
+			"cost": {
+				"input": 0.95,
+				"output": 3,
+				"cacheRead": 0.18,
+				"cacheWrite": 0
+			},
+			"contextWindow": 1048576,
+			"maxTokens": 32768,
+			"thinking": {
+				"mode": "effort",
+				"minLevel": "minimal",
+				"maxLevel": "xhigh"
+			}
+		}
+	},
 	"firepass": {
 		"kimi-k2.6-turbo": {
 			"id": "kimi-k2.6-turbo",

package/src/provider-models/descriptors.ts

@@ -13,6 +13,7 @@ import {
 	anthropicModelManagerOptions,
 	cerebrasModelManagerOptions,
 	cloudflareAiGatewayModelManagerOptions,
+	deepinfraModelManagerOptions,
 	deepseekModelManagerOptions,
 	firepassModelManagerOptions,
 	fireworksModelManagerOptions,
@@ -127,7 +128,7 @@ function catalogDescriptor(
  * OpenAI code provider) are handled separately because they require different config shapes.
  */
 export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = [
-	descriptor("anthropic", "claude-sonnet-4-6", config => anthropicModelManagerOptions(config)),
+	descriptor("anthropic", "claude-sonnet-5", config => anthropicModelManagerOptions(config)),
 	catalogDescriptor(
 		"alibaba-coding-plan",
 		"qwen3.5-plus",
@@ -168,6 +169,12 @@ export const PROVIDER_DESCRIPTORS: readonly ProviderDescriptor[] = [
 		config => deepseekModelManagerOptions(config),
 		catalog("DeepSeek", ["DEEPSEEK_API_KEY"]),
 	),
+	catalogDescriptor(
+		"deepinfra",
+		"deepseek-ai/DeepSeek-V3.2",
+		config => deepinfraModelManagerOptions(config),
+		catalog("DeepInfra", ["DEEPINFRA_API_KEY"]),
+	),
 	descriptor("mistral", "devstral-medium-latest", config => mistralModelManagerOptions(config)),
 	catalogDescriptor(
 		"nvidia",

package/src/provider-models/openai-compat.ts

@@ -665,6 +665,17 @@ export function deepseekModelManagerOptions(
 ): ModelManagerOptions<"openai-completions"> {
 	return createSimpleOpenAICompletionsOptions("deepseek", "https://api.deepseek.com", config);
 }
+
+export interface DeepInfraModelManagerConfig {
+	apiKey?: string;
+	baseUrl?: string;
+}
+
+export function deepinfraModelManagerOptions(
+	config?: DeepInfraModelManagerConfig,
+): ModelManagerOptions<"openai-completions"> {
+	return createSimpleOpenAICompletionsOptions("deepinfra", "https://api.deepinfra.com/v1/openai", config);
+}
 // ---------------------------------------------------------------------------
 // 7.5 Fireworks
 // ---------------------------------------------------------------------------
@@ -1709,7 +1720,7 @@ export interface GithubCopilotModelManagerConfig {
 }
 
 function inferCopilotApi(modelId: string): Api {
-	if (/^claude-(haiku|sonnet|opus)-4([.-]|$)/.test(modelId)) {
+	if (/^claude-(haiku|sonnet|opus)-(?:4|5)([.-]|$)/.test(modelId)) {
 		return "anthropic-messages";
 	}
 	if (modelId.startsWith("gpt-5") || modelId.startsWith("oswe")) {
@@ -2138,7 +2149,7 @@ const COPILOT_DEFAULT_RESOLUTION = {
 
 const COPILOT_API_RESOLUTION_RULES: readonly ApiResolutionRule[] = [
 	{
-		matches: modelId => /^claude-(haiku|sonnet|opus)-4([.-]|$)/.test(modelId),
+		matches: modelId => /^claude-(haiku|sonnet|opus)-(?:4|5)([.-]|$)/.test(modelId),
 		resolved: { api: "anthropic-messages", baseUrl: COPILOT_BASE_URL },
 	},
 	{
@@ -2282,6 +2293,8 @@ const MODELS_DEV_PROVIDER_DESCRIPTORS_CORE: readonly ModelsDevProviderDescriptor
 			requiresAssistantContentForToolCalls: true,
 		},
 	}),
+	// --- DeepInfra ---
+	openAiCompletionsDescriptor("deepinfra", "deepinfra", "https://api.deepinfra.com/v1/openai"),
 ];
 
 const MODELS_DEV_PROVIDER_DESCRIPTORS_CODING_PLANS: readonly ModelsDevProviderDescriptor[] = [

package/src/providers/anthropic.ts

@@ -21,6 +21,7 @@ import {
 } from "@gajae-code/utils";
 import { hasOpus47ApiRestrictions, mapEffortToAnthropicAdaptiveEffort } from "../model-thinking";
 import { calculateCost } from "../models";
+import { isUsageLimitError } from "../rate-limit-utils";
 import { getEnvApiKey, OUTPUT_FALLBACK_BUFFER } from "../stream";
 import type {
 	Api,
@@ -62,6 +63,7 @@ import { parseJsonWithRepair, parseStreamingJson } from "../utils/json-parse";
 import { parseGitHubCopilotApiKey } from "../utils/oauth/github-copilot";
 import { notifyProviderResponse } from "../utils/provider-response";
 import { isCopilotTransientModelError } from "../utils/retry";
+import { getRetryAfterMsFromHeaders } from "../utils/retry-after";
 import { resolveRetryBudget } from "../utils/retry-budget";
 import {
 	COMBINATOR_KEYS,
@@ -724,6 +726,7 @@ export type AnthropicClientOptionsArgs = {
 	onSseEvent?: AnthropicOptions["onSseEvent"];
 	fetch?: FetchImpl;
 	requestMaxRetries?: number;
+	maxRetryDelayMs?: number;
 };
 
 export type AnthropicClientOptionsResult = {
@@ -877,6 +880,101 @@ function mergeHeaders(...headerSources: (Record<string, string> | undefined)[]):
 	return merged;
 }
 
+const ANTHROPIC_RETRY_DELAY_CAP_MS = 60_000;
+const ANTHROPIC_RATE_LIMIT_HEADER_PREFIX = "anthropic-ratelimit-";
+
+function getSafeAnthropicHeaderEvidence(headers: Headers): string[] {
+	const evidence: string[] = [];
+	for (const [name, value] of headers) {
+		const lowerName = name.toLowerCase();
+		if (
+			lowerName === "retry-after" ||
+			lowerName === "retry-after-ms" ||
+			lowerName.startsWith(ANTHROPIC_RATE_LIMIT_HEADER_PREFIX)
+		) {
+			evidence.push(`${lowerName}=${value}`);
+		}
+	}
+	return evidence.sort();
+}
+
+function getStringProperty(source: unknown, key: string): string | undefined {
+	if (!isRecord(source)) return undefined;
+	const value = source[key];
+	return typeof value === "string" ? value : undefined;
+}
+
+function appendAnthropicRateLimitEvidence(bodyText: string, headers: Headers): string {
+	const evidence = getSafeAnthropicHeaderEvidence(headers);
+	if (evidence.length === 0) return bodyText;
+
+	const suffix = ` Anthropic rate-limit evidence: ${evidence.join(", ")}`;
+	try {
+		const parsed = JSON.parse(bodyText) as unknown;
+		if (isRecord(parsed)) {
+			const error = parsed.error;
+			if (isRecord(error) && typeof error.message === "string" && !error.message.includes(suffix)) {
+				return JSON.stringify({ ...parsed, error: { ...error, message: `${error.message}${suffix}` } });
+			}
+			if (typeof parsed.message === "string" && !parsed.message.includes(suffix)) {
+				return JSON.stringify({ ...parsed, message: `${parsed.message}${suffix}` });
+			}
+		}
+	} catch {}
+
+	return bodyText.includes(suffix) ? bodyText : `${bodyText}${suffix}`;
+}
+
+function isAnthropicUsageExhaustionResponse(
+	bodyText: string,
+	headers: Headers,
+	retryAfterMs: number | undefined,
+	retryDelayCapMs: number,
+): boolean {
+	const overageReason = headers.get("anthropic-ratelimit-unified-overage-disabled-reason")?.toLowerCase();
+	if (overageReason === "out_of_credits") return true;
+	if (retryAfterMs !== undefined && retryAfterMs > retryDelayCapMs) return true;
+
+	try {
+		const parsed = JSON.parse(bodyText) as unknown;
+		const error = isRecord(parsed) ? parsed.error : undefined;
+		const type = getStringProperty(error, "type") ?? getStringProperty(parsed, "type");
+		const message = getStringProperty(error, "message") ?? getStringProperty(parsed, "message") ?? bodyText;
+		return (
+			/rate_limit_error/i.test(type ?? "") &&
+			(/request would exceed your account.?s rate limit/i.test(message) || /out_of_credits/i.test(message))
+		);
+	} catch {
+		return /request would exceed your account.?s rate limit|out_of_credits/i.test(bodyText);
+	}
+}
+
+function wrapAnthropicFetchForBoundedRateLimits(baseFetch: FetchImpl, maxRetryDelayMs: number | undefined): FetchImpl {
+	const retryDelayCapMs = maxRetryDelayMs ?? ANTHROPIC_RETRY_DELAY_CAP_MS;
+	return Object.assign(
+		async (input: string | URL | Request, init?: RequestInit): Promise<Response> => {
+			const response = await baseFetch(input, init);
+			if (response.status !== 429 || retryDelayCapMs === 0) return response;
+
+			const headers = new Headers(response.headers);
+			const retryAfterMs = getRetryAfterMsFromHeaders(headers);
+			const bodyText = await response
+				.clone()
+				.text()
+				.catch(() => "");
+			if (!isAnthropicUsageExhaustionResponse(bodyText, headers, retryAfterMs, retryDelayCapMs)) return response;
+
+			headers.set("x-should-retry", "false");
+			return new Response(appendAnthropicRateLimitEvidence(bodyText, headers), {
+				status: response.status,
+				statusText: response.statusText,
+				headers,
+			});
+		},
+		baseFetch.preconnect ? { preconnect: baseFetch.preconnect } : {},
+	);
+}
+
 // The Anthropic SDK logs malformed SSE frames directly before rethrowing them.
 // We surface the resulting provider error ourselves, so keep the SDK quiet.
 const ANTHROPIC_SDK_LOG_LEVEL = "off" as const;
@@ -1042,6 +1140,7 @@ export function isProviderRetryableError(error: unknown, provider?: string): boo
 	if (!(error instanceof Error)) return false;
 	if (provider === "github-copilot" && isCopilotTransientModelError(error)) return true;
 	const msg = error.message.toLowerCase();
+	if (isUsageLimitError(error.message)) return false;
 	if (
 		isUnexpectedSocketCloseMessage(msg) ||
 		/rate.?limit|too many requests|overloaded|service.?unavailable|internal_error|stream error.*received from peer|1302|timed?\s*out while waiting for the first event|timeout waiting for first/i.test(
@@ -1166,6 +1265,7 @@ export const streamAnthropic: StreamFunction<"anthropic-messages"> = (
 					onSseEvent: options?.onSseEvent,
 					fetch: options?.fetch,
 					requestMaxRetries: options?.requestMaxRetries,
+					maxRetryDelayMs: options?.maxRetryDelayMs,
 				});
 				client = created.client;
 				isOAuthToken = created.isOAuthToken;
@@ -1723,11 +1823,8 @@ export function buildAnthropicClientOptions(args: AnthropicClientOptionsArgs): A
 	const foundryCustomHeaders = resolveAnthropicCustomHeaders(model);
 	const tlsFetchOptions = buildClaudeCodeTlsFetchOptions(model, baseUrl);
 	const baseFetch = args.fetch ?? fetch;
-	const debugFetch = onSseEvent
-		? wrapFetchForSseDebug(baseFetch, event => onSseEvent(event, model))
-		: args.fetch
-			? baseFetch
-			: undefined;
+	const boundedFetch = wrapAnthropicFetchForBoundedRateLimits(baseFetch, args.maxRetryDelayMs);
+	const debugFetch = onSseEvent ? wrapFetchForSseDebug(boundedFetch, event => onSseEvent(event, model)) : boundedFetch;
 	if (model.provider === "github-copilot") {
 		const copilotApiKey = parseGitHubCopilotApiKey(apiKey).accessToken;
 		const betaFeatures = [...extraBetas];
@@ -1755,7 +1852,7 @@ export function buildAnthropicClientOptions(args: AnthropicClientOptionsArgs): A
 			dangerouslyAllowBrowser: true,
 			defaultHeaders,
 			logLevel: ANTHROPIC_SDK_LOG_LEVEL,
-			...(debugFetch ? { fetch: debugFetch } : {}),
+			fetch: debugFetch,
 			...(tlsFetchOptions ? { fetchOptions: tlsFetchOptions } : {}),
 		};
 	}
@@ -1789,7 +1886,7 @@ export function buildAnthropicClientOptions(args: AnthropicClientOptionsArgs): A
 			dangerouslyAllowBrowser: true,
 			defaultHeaders,
 			logLevel: ANTHROPIC_SDK_LOG_LEVEL,
-			...(debugFetch ? { fetch: debugFetch } : {}),
+			fetch: debugFetch,
 		};
 	}
 
@@ -1802,7 +1899,7 @@ export function buildAnthropicClientOptions(args: AnthropicClientOptionsArgs): A
 		dangerouslyAllowBrowser: true,
 		defaultHeaders,
 		logLevel: ANTHROPIC_SDK_LOG_LEVEL,
-		...(debugFetch ? { fetch: debugFetch } : {}),
+		fetch: debugFetch,
 		...(tlsFetchOptions ? { fetchOptions: tlsFetchOptions } : {}),
 	};
 }

package/src/providers/google-gemini-headers.ts

@@ -3,8 +3,13 @@
  * Uses the same format as the official Gemini CLI (v0.35+):
  * GeminiCLI/VERSION/MODEL (PLATFORM; ARCH; SURFACE)
  */
+export const GEMINI_CLI_VERSION_ENV = "GJC_AI_GEMINI_CLI_VERSION";
+export const LEGACY_GEMINI_CLI_VERSION_ENV = "PI_AI_GEMINI_CLI_VERSION";
+export const DEFAULT_GEMINI_CLI_VERSION = "0.49.0";
+
 export function getGeminiCliUserAgent(modelId = "gemini-3.1-pro-preview"): string {
-	const version = process.env.PI_AI_GEMINI_CLI_VERSION || "0.46.0";
+	const version =
+		process.env[GEMINI_CLI_VERSION_ENV] || process.env[LEGACY_GEMINI_CLI_VERSION_ENV] || DEFAULT_GEMINI_CLI_VERSION;
 	const platform = process.platform === "win32" ? "win32" : process.platform;
 	const arch = process.arch === "x64" ? "x64" : process.arch;
 	return `GeminiCLI/${version}/${modelId} (${platform}; ${arch}; terminal)`;

package/src/providers/openai-completions.ts

@@ -1355,9 +1355,12 @@ export function parseChunkUsage(
 ): AssistantMessage["usage"] {
 	const promptTokenDetails = getOptionalObjectProperty(rawUsage, "prompt_tokens_details");
 	const completionTokenDetails = getOptionalObjectProperty(rawUsage, "completion_tokens_details");
+	const deepSeekCacheHitTokens = getOptionalNumberProperty(rawUsage, "prompt_cache_hit_tokens");
+	const deepSeekCacheMissTokens = getOptionalNumberProperty(rawUsage, "prompt_cache_miss_tokens");
 	const cachedTokens =
 		getOptionalNumberProperty(rawUsage, "cached_tokens") ??
 		(promptTokenDetails ? getOptionalNumberProperty(promptTokenDetails, "cached_tokens") : undefined) ??
+		deepSeekCacheHitTokens ??
 		0;
 	// OpenRouter exposes cache writes via `prompt_tokens_details.cache_write_tokens`
 	// and INCLUDES them in `prompt_tokens`. Without subtracting, cache-write tokens
@@ -1369,7 +1372,7 @@ export function parseChunkUsage(
 	const reasoningTokens =
 		(completionTokenDetails ? getOptionalNumberProperty(completionTokenDetails, "reasoning_tokens") : undefined) ?? 0;
 	const promptTokens = getOptionalNumberProperty(rawUsage, "prompt_tokens") ?? 0;
-	const input = Math.max(0, promptTokens - cachedTokens - cacheWriteTokens);
+	const input = Math.max(0, deepSeekCacheMissTokens ?? promptTokens - cachedTokens - cacheWriteTokens);
 	// Per OpenAI's CompletionUsage spec, `reasoning_tokens` is a subset of
 	// `completion_tokens` (which is the total billed output). Adding them would
 	// double-count.

package/src/rate-limit-utils.ts

@@ -36,6 +36,14 @@ export function parseRateLimitReason(errorMessage: string): RateLimitReason {
 		return "MODEL_CAPACITY_EXHAUSTED";
 	}
 
+	if (
+		lower.includes("out_of_credits") ||
+		lower.includes("request would exceed your account's rate limit") ||
+		lower.includes("request would exceed your accounts rate limit")
+	) {
+		return "QUOTA_EXHAUSTED";
+	}
+
 	if (
 		lower.includes("per minute") ||
 		lower.includes("rate limit") ||
@@ -86,8 +94,7 @@ export function calculateRateLimitBackoffMs(reason: RateLimitReason): number {
 
 /** Detect usage/quota limit errors in error messages (persistent, requires credential switch). */
 const USAGE_LIMIT_PATTERN =
-	/usage.?limit|usage_limit_reached|usage_not_included|limit_reached|model.?limit|model_limit_reached|message.?limit|message_limit_reached|limit for this model|quota.?exceeded|resource has been exhausted[^\n]*(?:quota|limit)/i;
-
+	/usage.?limit|usage_limit_reached|usage_not_included|limit_reached|model.?limit|model_limit_reached|message.?limit|message_limit_reached|limit for this model|quota.?exceeded|out_of_credits|request would exceed your account.?s rate limit|resource has been exhausted[^\n]*(?:quota|limit)/i;
 export function isUsageLimitError(errorMessage: string): boolean {
 	return USAGE_LIMIT_PATTERN.test(errorMessage);
 }

package/src/stream.ts

@@ -98,6 +98,7 @@ const serviceProviderMap: Record<string, KeyResolver> = {
 	"opencode-zen": "OPENCODE_API_KEY",
 	cursor: "CURSOR_ACCESS_TOKEN",
 	deepseek: "DEEPSEEK_API_KEY",
+	deepinfra: "DEEPINFRA_API_KEY",
 	"openai-codex": "OPENAI_CODEX_OAUTH_TOKEN",
 	"azure-openai": "AZURE_OPENAI_API_KEY",
 	"azure-openai-responses": "AZURE_OPENAI_API_KEY",

package/src/types.ts

@@ -116,6 +116,7 @@ export type KnownProvider =
 	| "gitlab-duo"
 	| "cursor"
 	| "deepseek"
+	| "deepinfra"
 	| "xai"
 	| "groq"
 	| "cerebras"
@@ -186,7 +187,7 @@ export type CacheRetention = "none" | "short" | "long";
  *
  * The unscoped values (`"auto"`, `"default"`, `"flex"`, `"scale"`,
  * `"priority"`) are passed through to providers that understand them
- * (OpenAI's `service_tier` field directly; Anthropic translates
+ * (OpenAI and DeepInfra's `service_tier` field directly; Anthropic translates
  * `"priority"` into `speed: "fast"` on supported Opus models).
  *
  * The scoped values target a specific provider family and behave as the
@@ -223,17 +224,17 @@ export function resolveServiceTier(
 }
 
 /**
- * True when the (possibly scoped) tier should be sent as OpenAI's
- * `service_tier` request field for the given provider. Non-OpenAI
- * providers, unsupported tiers (`"auto"`, `"default"`), and scope
- * mismatches all return false.
+ * True when the (possibly scoped) tier should be sent as an OpenAI-compatible
+ * `service_tier` request field for providers that support it. Unsupported tiers
+ * (`"auto"`, `"default"`) and scope mismatches all return false.
  */
 export function shouldSendServiceTier(
 	serviceTier: ServiceTier | null | undefined,
 	provider: Provider | undefined,
 ): boolean {
-	if (provider !== "openai" && provider !== "openai-codex") return false;
 	const resolved = resolveServiceTier(serviceTier, provider);
+	if (provider === "deepinfra") return resolved === "priority";
+	if (provider !== "openai" && provider !== "openai-codex") return false;
 	return resolved === "flex" || resolved === "scale" || resolved === "priority";
 }
 
@@ -252,7 +253,9 @@ export function getPriorityPremiumRequests(
 	if (resolveServiceTier(serviceTier, provider) !== "priority") return 0;
 	// Only providers that realize `priority` on the wire bill the user.
 	// Everywhere else, the field is silently dropped and nothing is charged.
-	return provider === "openai" || provider === "openai-codex" || provider === "anthropic" ? 1 : 0;
+	return provider === "openai" || provider === "openai-codex" || provider === "anthropic" || provider === "deepinfra"
+		? 1
+		: 0;
 }
 
 export interface ProviderSessionState {

package/src/utils/oauth/deepinfra.ts

@@ -0,0 +1,15 @@
+/** DeepInfra login flow (API key paste against https://api.deepinfra.com/v1/openai). */
+import { createApiKeyLogin } from "./api-key-login";
+
+export const loginDeepInfra = createApiKeyLogin({
+	providerLabel: "DeepInfra",
+	authUrl: "https://deepinfra.com/dash/api_keys",
+	instructions: "Create or copy your DeepInfra API key from the DeepInfra dashboard",
+	promptMessage: "Paste your DeepInfra API key",
+	placeholder: "sk-...",
+	validation: {
+		kind: "models-endpoint",
+		provider: "DeepInfra",
+		modelsUrl: "https://api.deepinfra.com/v1/openai/models",
+	},
+});

package/src/utils/oauth/index.ts

@@ -60,6 +60,11 @@ const builtInOAuthProviders: OAuthProviderInfo[] = [
 		name: "DeepSeek",
 		available: true,
 	},
+	{
+		id: "deepinfra",
+		name: "DeepInfra",
+		available: true,
+	},
 	{
 		id: "xai",
 		name: "xAI",
@@ -356,6 +361,7 @@ export async function refreshOAuthToken(
 		case "opencode-zen":
 		case "opencode-go":
 		case "cerebras":
+		case "deepinfra":
 		case "fireworks":
 		case "firepass":
 		case "fugu":

package/src/utils/oauth/types.ts

@@ -15,6 +15,7 @@ export type OAuthProvider =
 	| "cloudflare-ai-gateway"
 	| "cursor"
 	| "deepseek"
+	| "deepinfra"
 	| "fireworks"
 	| "firepass"
 	| "fugu"