The Short List
| Model | Provider | Context | Best For | Listed Limit |
|---|---|---|---|---|
| Gemini 2.5 Flash | Google AI Studio | 1,000K | All-round, long docs | 15 RPM, 1,500 RPD |
| Moonshot Kimi K2 | Groq | 131K | Speed, chat, coding | Check console |
| Nemotron 3 Super | NVIDIA NIM | 262K | Reasoning, math, coding | Check provider |
| Codestral | Mistral AI | 256K | Code generation | ~1 RPS, 500K TPM |
| GPT-OSS-120B | Cerebras | 128K model / 8K free tier | Heavy workloads, big model | 30 RPM, 1M TPD |
1. Google Gemini 2.5 Flash — The All-Rounder
Provider: Google AI Studio · Context: 1,000,000 tokens · Rate: 15 RPM, 1,500 RPD
Google lists a 1 million token context window for Gemini 2.5 Flash. It accepts text, image, audio, and video input, making it a practical option for long and multimodal prompts.
Key Highlights
- 1M context window
- Multimodal (text + image + audio + video)
- 15 RPM, 1,500 RPD
- Long-document and codebase workflows
ANTHROPIC_BASE_URL="https://generativelanguage.googleapis.com/v1beta"
ANTHROPIC_AUTH_TOKEN="YOUR_GOOGLE_API_KEY"Account and quota requirements can change. Check Google AI Studio and get an API key →
2. Groq Moonshot Kimi K2 — The Low-Latency Option
Provider: Groq · Context: 131,072 tokens · Rate: Check the current Groq console
Groq's LPU infrastructure is designed for low-latency inference. Its current catalog listing for Moonshot Kimi K2 is a general-purpose option with a 131K context window.
Key Highlights
- Low-latency LPU inference
- 131K context
- Current quota shown in the Groq console
- Chat and coding workflows
ANTHROPIC_BASE_URL="https://api.groq.com/openai/v1"
ANTHROPIC_AUTH_TOKEN="YOUR_GROQ_API_KEY"3. NVIDIA Nemotron 3 Super — The Reasoner
Provider: NVIDIA NIM · Context: 262,000 tokens · Rate: Free tier
NVIDIA's own model with 262K context and strong reasoning capabilities. Free tier via NVIDIA NIM with solid rate limits. Great for technical reasoning, math, and coding.
Key Highlights
- 262K context window
- Strong reasoning and math
- OpenAI-compatible API
- Free tier with generous limits
ANTHROPIC_BASE_URL="https://integrate.api.nvidia.com/v1"
ANTHROPIC_AUTH_TOKEN="YOUR_NVIDIA_API_KEY"4. Mistral Codestral — The Coder
Provider: Mistral AI · Context: 256,000 tokens · Rate: ~1 RPS, 500K TPM
Mistral positions Codestral as a coding model for code generation, debugging, and refactoring. This catalog listing shows a 256K context window.
Key Highlights
- Purpose-built for coding
- 256K context
- ~1 RPS, 500K TPM
- Supports FIM (fill-in-the-middle)
ANTHROPIC_BASE_URL="https://api.mistral.ai/v1"
ANTHROPIC_AUTH_TOKEN="YOUR_MISTRAL_API_KEY"5. Cerebras GPT-OSS-120B — The Heavy Lifter
Provider: Cerebras · Context: 128,000 model maximum; the listed free tier is limited to 8K · Rate: 30 RPM, 14,400 RPD, 1M TPD
Cerebras hosts this 120B-parameter open model with a listed free tier. Check the current request and daily token limits before using it for sustained workloads.
Key Highlights
- 120B parameters
- 128K model context (free tier may be lower)
- 30 RPM, 1M TPD
- Open-weight model
ANTHROPIC_BASE_URL="https://api.cerebras.ai/v1"
ANTHROPIC_AUTH_TOKEN="YOUR_CEREBRAS_API_KEY"How We Picked These 5
We currently list 307 available free-tier models from 25 providers. The catalog changes as providers add, remove, or restrict endpoints. These 5 were selected based on:
- Context window — all above 128K, enough for real work
- Rate limits — practical for daily use (not 1 RPM)
- Provider documentation — setup and limit claims can be checked against provider sources
- Account requirements — confirm current signup, billing, and quota rules before integrating
- Use case coverage — coding, reasoning, chat, and long documents
Data Notes & Official Sources
Limits were reviewed on . Providers can change free-tier quotas, model access, account requirements, and regional availability without notice. Confirm the current values before building a production integration.
- Google Gemini API rate limits
- Groq rate limits
- NVIDIA API model catalog
- Mistral usage limits
- Cerebras rate limits
How to Use These with Claude Code (cc)
Set two environment variables and you're done:
# Pick your model's config from above
export ANTHROPIC_BASE_URL="https://api.groq.com/openai/v1"
export ANTHROPIC_AUTH_TOKEN="your-api-key"
# Then run cc normally — it now uses the free backend
claudeFor Cursor, OpenCode, and other tools, grab the one-click config from each model's detail page on freellm.net.
Browse all 307 available free-tier models → Model Directory