Groq retired its free Llama models — what to use instead

Last verified: 2026-10-10

On 2026-08-16 Groq removed llama-3.3-70b-versatile and llama-3.1-8b-instant from free and developer accounts. Code and tutorials that still use those ids get an error like this:

{
  "error": {
    "message": "The model `llama-3.3-70b-versatile` has been decommissioned and is no longer supported. ...",
    "type": "invalid_request_error",
    "code": "model_decommissioned"
  }
}

1. Change the model id (keep Groq)

Base URL and API key stay the same. Free limits on Groq are now 30 RPM and 1,000 requests/day.

curl https://api.groq.com/openai/v1/chat/completions \
  -H "Authorization: Bearer $GROQ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-oss-120b",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

2. Need Llama specifically? Use Cloudflare Workers AI

Cloudflare is the main free hosted Llama route left: @cf/meta/llama-3.3-70b-instruct-fp8-fast and @cf/meta/llama-3.1-8b-instruct-fast. 10,000 Neurons/day shared across all models — not a request count. Bigger models burn more Neurons per call.

curl https://api.cloudflare.com/client/v4/accounts/$CLOUDFLARE_ACCOUNT_ID/ai/run/@cf/meta/llama-3.3-70b-instruct-fp8-fast \
  -H "Authorization: Bearer $CLOUDFLARE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Hello"}]}'

Details: Llama 3.3 free API · Llama 3.1 free API · Cloudflare vs Groq

3. Or run Llama locally

Llama 3.1 8B on Ollama runs on an 8 GB GPU and exposes the same OpenAI-style API at http://localhost:11434/v1.

Update your editor config

• Cursor + Groq — now uses GPT-OSS
• VS Code + Continue + Groq
• Fallback across Groq, Gemini and OpenRouter

FAQ

Why does Groq say llama-3.3-70b-versatile is decommissioned?

Groq retired llama-3.3-70b-versatile and llama-3.1-8b-instant for free and developer accounts on 2026-08-16. Requests with those model ids now fail. Switch to openai/gpt-oss-120b, openai/gpt-oss-20b or qwen/qwen3.8-27b.

What replaces llama-3.1-8b-instant on Groq?

openai/gpt-oss-20b is the small, fast option on Groq's free tier. Same API, same key — change only the model string.

Can I still use Llama for free through an API?

Yes, on Cloudflare Workers AI: @cf/meta/llama-3.3-70b-instruct-fp8-fast and @cf/meta/llama-3.1-8b-instruct-fast, within 10,000 free Neurons per day. Or run Llama locally with Ollama.

Did Groq's free limits change too?

Yes. The free plan is now 30 requests per minute and 1,000 requests per day per model, down from the 14,400 requests/day many older tutorials quote.

← Free API guides · Groq free tier · GPT-OSS free API · Qwen3.8 free API