Groq

Custom LPU silicon serving open models at absurd token speeds, with a generous free tier — the fastest way to feel gpt-oss or Kimi.

Get an API key

1. Sign up at console.groq.com and verify your email
2. Open 'API Keys', click 'Create API Key'
3. Name the key and submit
4. Copy it immediately, then set GROQ_API_KEY

Create a key ↗ · Pricing ↗

Free tier, no credit card — all models with per-minute/daily limits.

Enable

export GROQ_API_KEY=...        # shell profile — survives forever

or in the terminal — validates the key live before saving:

/enable groq

or pick any locked 🔒 Groq model in /models and paste the key inline.

Models

Starter menu (once enabled, /models fetches the provider's full live list):

• openai/gpt-oss-120b
• openai/gpt-oss-20b
• moonshotai/kimi-k2-instruct-0905
• qwen/qwen3-32b
• llama-3.3-70b-versatile

Switch anytime — /model gpt-oss-120b fuzzy-matches; context carries over.

SDK

from mantis_agent import MantisAgentOptions
options = MantisAgentOptions(
    model="openai/gpt-oss-120b",
    backend="https://api.groq.com/openai/v1",     # key read from $GROQ_API_KEY
)

Notes

• Free tier is rate-limited but real — great for evaluation.
• Ids are namespaced (openai/gpt-oss-120b, qwen/qwen3-32b).