it's free*.ai

目录 · 核实于 2 Sep 2026

说 OpenAI 协议的免费 API

这里的每一家都接受与 OpenAI SDK 相同的请求格式。把 `baseURL` 指向它们的端点、换上密钥、保留其余代码 —— 同样两行改动适用于 TypeScript 和 Python 官方 SDK，也适用于 Cursor、Claude Code、Aider 和 Cline。

这让它们在实践中可以互换：某家免费额度用尽时，直接换到表中的下一行，应用代码一行不用改。

服务商免费额度上下文模型数注册要求包含模型
Google AI Studiomidudev 推荐每天 1,500 次请求1M17 个模型
Gemini 3.7 FlashGemini 3.6 FlashGemini 3.5 FlashGemini 3.5 Flash LiteGemini 3.1 Flash LiteGemini 3 FlashGemini 2.5 FlashGemini 2.5 Flash LiteGemini 2.5 Flash ImageGemini 2.0 FlashGemini 2.0 Flash LiteGemini 3.1 Flash ImageGemini 3.1 Flash TTSGemini 2.5 Flash TTSGemini Embedding 001Gemini Embedding 2Imagen 4 Fast
Groqmidudev 推荐每天 1,000 次请求131K7 个模型
GPT-OSS 120BQwen 3.8 27BGPT-OSS 20BGPT-OSS Safeguard 20BQwen 3.6 27BGroq CompoundGroq Compound Mini
NVIDIA NIMmidudev 推荐每分钟 40 次请求1M82 个模型
GPT-OSS 120BYi LargeFuyu 8BJamba 1.5 LargeSea Lion 7BStarcoder 2 15BDbrxDeepSeek Coder 6.7BDeepSeek V4 Flash 0731DeepSeek V4 Pro 0813Codegemma 1.1 7BCodegemma 7BDeplotDiffusiongemma 26B A4BGemma 2BGemma 3 12BGemma 3 4BGemma 4 31BRecurrentgemma 2BGranite 3.0 3B A800mGranite 3.0 8BGranite 34B CodeGranite 8B CodeCodellama 70BLlama 3.2 11B VisionLlama 3.2 90B VisionLlama Guard 4 12BLlama 2 70BMuse Glimmer 30BKosmos 2Phi 3 Vision 128kPhi 3.5 MoEMiniMax M3Codestral 22B V0.1Mistral 7B V0.3Mistral LargeMistral Large 2Mistral NemotronMixtral 8x22B V0.1Kimi K2.6Kimi K3Mistral Nemo 12BAi Synthetic Video DetectorCosmos Reason 2 8BEmbed Qa 4Ising Calibration 1.5 31BLlama 3.1 Nemoguard 8B Content SafetyLlama 3.1 Nemoguard 8B Topic ControlLlama 3.1 Nemotron 51BLlama 3.1 Nemotron 70BLlama 3.1 Nemotron Safety Guard 8B V3Llama 3.1 Nemotron Ultra 253B V1Llama 3.2 Nemoretriever 1B Vlm Embed V1Llama 3.2 Nv Embedqa 1B V1Llama Nemotron Embed VL 1B V2Llama 3 Chatqa 1.5 70BMistral Nemo Minitron 8B 8kNemotron 3 Embed 1BNemotron 3 Nano Omni 30B A3B ReasoningNemotron 3 Super 120B A12BNemotron 3 Ultra 550B A55BNemotron 3.5 Content SafetyNemotron 3.5 Lightning 30B A3BNemotron 4 340BNemotron 4 340B RewardNemotron Nano 3 30B A3BNemotron ParseNeva 22BNv Embedqa Mistral 7B V2NvclipRiva Translate 4BRiva Translate 4B V1.1Riva Translate 4B V2VilaGPT-OSS 20BLaguna XS 2.1Arctic Embed LPalmyra Creative 122BPalmyra Fin 70B 32kPalmyra Med 70BPalmyra Med 70B 32kZamba 2 7B
OpenRoutermidudev 推荐18 个模型零成本1M18 个模型
Ling 3.0 Flash FinDots 3 Note PreviewLFM 2.5 2.6BNemotron 3.5 LightningInkling SmallLaguna S 2.1InklingLaguna XS 2.1North Mini CodeGLM 5.2Nemotron 3.5 Content SafetyNemotron 3 UltraMiniMax M3Nemotron 3 Nano OmniGemma 4 26B A4BGemma 4 31BMiniMax M2.7Nemotron 3 SuperFree Models Router
Vercel AI Gatewaymidudev 推荐每月 $5 赠款1M12 个模型
GPT-OSS 120BGPT 5.4 MiniGemma 4 31BGemini 2.5 FlashQwen 3.8 Flash NextGLM 5.3 FlashKimi K2.7 CodeMiniMax M3Nemotron 3 SuperLlama 4 MaverickGrok 4.6DeepSeek V3.2 Thinking
AMD Radeon Cloud每天 $10 用量1M5 个模型
DeepSeek V4 FlashDeepSeek V4 Flash VisionQwen 3.8 Flash NextMiniCPM V 4.6MiniCPM 5 1B
Requesty每天 200 次请求1M12 个模型
Laguna M.1Laguna XS.2Ling 3.0 TinyGemma 4 31BLeanstral 1.5Nemotron 3 Super 120B A12BNemotron 3 Nano Omni 30B A3B ReasoningNemotron 3 Nano 30B A3BNemotron 3.5 Content SafetyNemotron 3 Ultra 550B A55BMuse Glimmer 30BNemotron 3.5 Lightning 30B A3B
Cerebras一次性 $5 赠款65K2 个模型
GPT-OSS 120BGemma 4 31B
Cloudflare Workers AI每天 10,000 神经元262K40 个模型
GPT-OSS 120BQwen 3.8 27BGPT-OSS 20BLlama 3.3 70BLlama 4 ScoutLlama 3.1 8B FastLlama 3.1 8B FP8Llama 3.2 11B VisionLlama 3.2 1BLlama 3.2 3BGemma 4 26BGLM 4.7 FlashGranite 4.0 H MicroNemotron 3 SuperQwen 3 30B A3BQwen 2.5 Coder 32BQwQ 32BMistral Small 3.1DeepSeek R1 Distill 32BLlama Guard 3 8BWhisper Large V3 TurboBGE M3BGE BaseEmbeddingGemma 300MFLUX.1 SchnellQwen 3 Embedding 0.6BGemma 3 12BLlama 3.1 8BMistral 7BLLaVA 1.5 7BWhisperNova 3MeloTTSStable Diffusion XLResnet 50BART Large CNNDistilBERT SST 2SQLCoder 7BPhi 2
Mistral AIFree mode 与 Labs256K12 个模型
Mistral SmallMistral Small 2506Mistral NemoCodestralPixtral 12BLabs Mistral SmallMistral EmbedMistral ModerationMistral OCRMistral Small 3.1Devstral SmallMagistral Small
Cohere每月 1,000 次请求256K12 个模型
Command ACommand A+Command R+Command RCommand R7BCommand A ReasoningCommand A TranslateEmbed V4Embed English V3Embed Multilingual V3Rerank V3.5Rerank English V3
LLM7.io每分钟 10 次（带 token 40 次）1M5 个模型
GPT-OSS 120BGemma 4 31BMiniMax M2.7CodestralMistral Nemo
DeepSeek注册送赠款1M3 个模型
DeepSeek V4 FlashDeepSeek V4 ProDeepSeek V4 Flash Vision
Hugging Face每月 $0.10 赠款131K135 个模型
GPT-OSS 120BQwen 3.8 27BGLM 5.3 FlashGLM 5.3Kimi K3DeepSeek V4 Flash 0731Granite 4.2 30BGLM 5.3 Flash BF16Muse Glimmer 30BGemma 4 31BGranite 4.2 3BLlama 3.1 8BGranite 4.2 8BGLM 5.3 BF16Qwen 3.8 2.4T A95BDeepSeek V4 Pro 0813DeepSeek V4 FlashQwen 3.5 9BInkling SmallQwen 3.6 35B A3BGemma 4 26B A4BHy3GLM 5.2InklingLing 3.0 FlashGPT-OSS 20BMiniMax M3DeepSeek V4 ProTernary Bonsai 27B GgufGLM 4.7 FlashDeepSeek R1Qwen 3 8BLlama 3.3 70BQwen 2.5 Coder 7BNVIDIA Nemotron 3 Ultra 550B A55B NVFP4Qwen 3 Coder NextNVIDIA Nemotron 3.5 Lightning 30B A3B BF16Qwen 3 Coder 30B A3BQwen 3 4B 2507Qwen 3.5 35B A3BQwen 3.6 27BMiMo V2.5Qwen 2.5 Coder 32BApertus V1.5 8BQwen 3 14BKimi K2.6GLM 5.2 FP8Kimi K2.7 CodePhi 4Gemma 3 4BQwen 3 235B A22B 2507Apertus V1.5 70BL3 8B Stheno V3.2Qwen 3 30B A3BStep 3.7 FlashQwen 3 Next 80B A3BDeepSeek V3.2MiMo V2.5 ProLlama 4 Scout 17B 16EQwen 3 32BQwen 3.5 397B A17BQwen 3 235B A22BKimi K2.5Gemma 3 27BQwen 3.5 27BQwen 3.5 122B A10BQwen 3 235B A22B Thinking 2507Tiny Aya GlobalGLM 5Gemma 3 12BHermes 3 Llama 3.1 70BTiny Aya WaterGLM 5.1Llama Guard 4 12BQwen 2.5 72BGLM 5.1 FP8Apertus 8B 2509DeepSeek V3GLM 4.7DeepSeek V3 0324DeepSeek R1 Distill Llama 8BWizardLM 2 8x22BERNIE 4.5 VL 424B A47B Base PTDeepSeek V3.2 ExpGLM 4.6V FlashAutoGLM Phone 9B MultilingualTernary Bonsai 27B AWQ 4bitL3 8B Lunaris V1Gemma SEA LION V4 27BLlama 4 Maverick 17B 128E FP8Qwen 3 VL 235B A22BDeepSeek R1 Distill Llama 70BApertus 70B 2509GLM 4.5VQwen 3 VL 30B A3BGLM 4 32B 0414Qwen SEA LION V4 32BQwen 3 VL 235B A22B ThinkingQwen 2.5 Coder 3BMiniMax M1 80kBielik 11B V3.0Qwen 2.5 VL 72BDeepSeek R1 Distill Qwen 7BKimi K2DeepSeek R1 0528GLM 4.5DeepSeek R1 Distill Qwen 14BGLM 4.5 AirDeepSeek V3.1GLM 4.6 FP8Qwen 3 4B Thinking 2507Qwen 3 Coder 480B A35BDeepSeek V3.1 TerminusC4ai Command R 08 2024GLM 4.6Kimi K2 0905Step 3.5 FlashAya Expanse 32BGLM 4.7 FP8MiniMax M2MiniMax M2.7C4ai Command R7B 12 2024MiniMax M2.1GLM 4.5V FP8C4ai Command R7B Arabic 02 2025MiniMax M2.5GLM 4.6VC4ai Command A 03 2025Aya Vision 32BGLM 4.6V FP8GPT-OSS Safeguard 20BCommand A Reasoning 08 2025Command A Translate 08 2025Tiny Aya EarthTiny Aya Fire
ModelScope每天 2,000 次请求1M50 个模型
Qwen 3.8 27BDeepSeek V4 FlashDeepSeek V4 ProDeepSeek V4 Pro 0813EA 29B A4BAntAngelMedLongCat Flash LiteMiniMax M1MiniMax M3Mistral LargeQwen Image EditCompassJudger 32BInternVL 3.5ERNIE 4.5 0.3BERNIE 4.5 21BERNIE 4.5 300BERNIE 4.5 VL 28BQwen Image EditQwen 3 14BQwen 3 235BQwen 3 235BQwen 3 235B ThinkingQwen 3 30B A3BQwen 3 30B A3B ThinkingQwen 3 4BQwen 3 8BQwen 3 Coder 30BQwen 3 Embedding 0.6BQwen 3 Embedding 4BQwen 3 Embedding 8BQwen 3 Next 80BQwen 3 Next 80B ThinkingQwen 3 VL 235BQwen 3 VL 8BQwen 3 VL 8B ThinkingQwen 3.5 122BQwen 3.5 27BQwen 3.5 35BQwen 3.5 397BQwen 3.8 Flash NextIntern S1Intern S1 MiniIntern S2Step 3.5 FlashStep 3.7 FlashHunyuan Hy3XiYanSQL 32BXiYanSQL 32B 2504GLM 4.7 FlashGLM 5.2
Z.ai (Zhipu)GLM Flash 系列免费200K4 个模型
GLM 4.7 FlashGLM 4.5 FlashGLM 4.6 FlashGLM 4 Flash
SambaNova每天 20 次请求1M7 个模型
GPT-OSS 120BLlama 3.3 70BDeepSeek V3.1DeepSeek V3.2MiniMax M2.7MiniMax M3Gemma 4 31B
Ollama Cloud每月启动赠金1M19 个模型
GPT-OSS 120BGLM 5.2Kimi K2.6Kimi K3DeepSeek V4 Pro 0813MiniMax M3GLM 5.1GLM 5.3DeepSeek V4 Flash 0731GPT-OSS 20BMiniMax M2.7Mistral Large 3 675BNemotron 3 SuperNemotron 3 Nano 30BGemma 4 31BQwen 3.5 397BNemotron 3 UltraGLM 5.3 FlashKimi K2.7 Code
OVHcloud AI Endpoints7 个模型 €0262K7 个模型
Riva TTS EnglishRiva TTS GermanRiva TTS SpanishRiva TTS ItalianQwen 3Guard Gen 0.6BQwen 3Guard Gen 8BStable Diffusion XL
Alibaba Model Studio每模型 1M token，90 天1M— 个模型
Qwen 3.7 PlusQwen 3.6 PlusQwen 3.5 PlusQwen 3 Coder PlusQwen 3.5 Plus Long
OpenCode Zen6 个免费模型1M6 个模型
Nemotron 3 UltraNemotron 3 LightningMiMoLing FlashBig PickleMuse Spark
AI21 Labs3 个月 $10 赠款256K2 个模型
Jamba Mini 2Jamba Large 1.6
SiliconFlow实名认证后按模型限额131K3 个模型
Qwen 3 8BQwen 2.5 7BGLM 4 9B
Nebius AI Studio30 天 $1 试用128K1 个模型
Llama 3.3 70B
Nscale$5 启动赠金128K2 个模型
Llama 4 ScoutLlama 3.3 70B

关于这些 API 的常见问题

具体哪些东西必须一致？

/chat/completions 路径、messages 数组和响应外层结构。附加能力各有差异 —— 工具调用、结构化输出和视觉的支持程度参差，依赖前先看服务商页面。

能自动切换到别家吗？

可以，这正是兼容端点最大的好处。捕获 429、把客户端换到下一个接口地址再重试；代码其余部分完全无感。

换个角度看