Providers

• Pollinations

  Free, no key required

  NO KEY

  Community-run open API that serves open-weight models anonymously. The easiest way to try the playground: pick it and hit run.

  • No sign-up
  • GPT-OSS 20B
  • Rate limited

  Suggested models

  • openai-fast

  Test Pollinations API docs

• Groq

  LPU hardware, very low latency

  BYO KEY

  Custom Language Processing Units built for inference. Known for extremely high tokens-per-second on Llama and GPT-OSS models.

  • Free tier
  • Fast TTFT
  • OpenAI-compatible

  Suggested models

  • llama-3.3-70b-versatile
  • llama-3.1-8b-instant
  • openai/gpt-oss-120b
  • openai/gpt-oss-20b

  Test Groq API docs

• Together AI

  Broad open-model catalog

  BYO KEY

  Serverless endpoints for hundreds of open models, from Llama and Qwen to DeepSeek, plus fine-tuning and dedicated GPUs.

  • Huge catalog
  • Fine-tuning
  • OpenAI-compatible

  Suggested models

  • meta-llama/Llama-3.3-70B-Instruct-Turbo
  • Qwen/Qwen2.5-72B-Instruct-Turbo
  • deepseek-ai/DeepSeek-V3

  Test Together AI API docs

• OpenRouter

  One key, many upstreams

  BYO KEY

  A router in front of dozens of providers. Includes a rotating set of free open models (IDs ending in :free).

  • Free models
  • Fallback routing
  • Unified billing

  Suggested models

  • openai/gpt-oss-20b:free
  • meta-llama/llama-3.3-70b-instruct
  • qwen/qwen-2.5-72b-instruct

  Test OpenRouter API docs

• Hugging Face

  Inference Providers router

  BYO KEY

  Hugging Face's router forwards requests to partner providers for models hosted on the Hub, using your HF access token.

  • Hub models
  • Monthly free credits
  • Multi-provider

  Suggested models

  • meta-llama/Llama-3.1-8B-Instruct
  • Qwen/Qwen2.5-72B-Instruct
  • deepseek-ai/DeepSeek-V3-0324

  Test Hugging Face API docs

• DeepInfra

  Low-cost serverless GPUs

  BYO KEY

  Pay-per-token hosting for popular open models with an OpenAI-compatible endpoint.

  • Low prices
  • Many models
  • OpenAI-compatible

  Suggested models

  • meta-llama/Meta-Llama-3.1-8B-Instruct
  • meta-llama/Llama-3.3-70B-Instruct
  • Qwen/Qwen2.5-72B-Instruct

  Test DeepInfra API docs

• Fireworks AI

  Optimised serving stack

  BYO KEY

  Production inference for open models with its own serving engine and function-calling support.

  • Function calling
  • JSON mode
  • OpenAI-compatible

  Suggested models

  • accounts/fireworks/models/llama-v3p3-70b-instruct
  • accounts/fireworks/models/deepseek-v3
  • accounts/fireworks/models/qwen2p5-72b-instruct

  Test Fireworks AI API docs

• Cerebras

  Wafer-scale inference

  BYO KEY

  Runs open models on wafer-scale chips for some of the highest output speeds available.

  • Free tier
  • Very high tok/s
  • OpenAI-compatible

  Suggested models

  • llama3.1-8b
  • llama-3.3-70b
  • gpt-oss-120b

  Test Cerebras API docs

• Mistral AI

  Open-weight models from Mistral

  BYO KEY

  La Plateforme serves Mistral's own models, including the Apache-licensed open-weight releases.

  • Apache 2.0 weights
  • EU-hosted
  • Free tier

  Suggested models

  • open-mistral-nemo
  • mistral-small-latest

  Test Mistral AI API docs

Missing a provider? Any OpenAI-compatible endpoint can be added in lib/providers.ts. See the docs.