ModelHubby

← All guides · provider record →

Cohere — free tier setup

What you get free

A non-expiring trial key covering the Command family (chat, tool calling, vision) plus Embed and Rerank, no credit card. Catalog quota: "Official rate-limits docs: 'Trial keys (and prod keys on newer Chat model variants) are limited to 1,000 API calls a month.' Chat API (all models): 20 req / min. Rerank: 10 req / min. Embed: 2,000 inputs / min. cheahjs README: 'Models share a common monthly quota.'" Category: conditional_free — the condition: the trial key is for non-commercial use only (per the catalog / mnfst data.json; the official rate-limits page describes trial keys only as "free but limited in usage" without restating this). Do not ship a commercial product on a trial key. The other gotcha: 1,000 calls/month is a shared pool across all models — it goes fast.

Friction check
Credit card no
Phone verification unknown
Org/membership no
Trial key licence non-commercial use only
Est. time to first key 5m
Trains on your prompts unknown

Get a key

Exact button names unconfirmed (dashboard requires login) — flow per catalog:

  1. Sign up at dashboard.cohere.com (email or OAuth).
  2. Go to dashboard.cohere.com/api-keys.
  3. A free Trial key is available there — copy it (creation may be automatic on signup; unconfirmed — verify on first signup).
  4. No card, no plan selection needed for trial usage.

First request

Using the OpenAI-compatible Compatibility API (confirmed in official docs per catalog; native endpoint is https://api.cohere.com/v2). Success shape: JSON with choices[0].message.content.

export COHERE_API_KEY="YOUR_KEY"
curl https://api.cohere.ai/compatibility/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $COHERE_API_KEY" \
  -d '{"model":"command-r7b-12-2024","messages":[{"role":"user","content":"Say OK."}],"max_tokens":16}'

Limits that will bite you

  • 1,000 API calls a month, total, shared across all models (catalog: shared_pool true). At 20 RPM you can exhaust a month's quota in under an hour of sustained traffic.
  • Chat: 20 req/min across all Command models (command-a-03-2025, command-a-plus-05-2026, command-a-vision-07-2025, command-r-08-2024, command-r7b-12-2024).
  • Rerank: 10 req/min. Embed: 2,000 inputs/min on trial (official rate-limits page).
  • Max output ~4,000 tokens on Command chat models (catalog).
  • The monthly cap also applies to production keys on newer Chat model variants ("Trial keys (and prod keys on newer Chat model variants) are limited to 1,000 API calls a month") — upgrading doesn't automatically remove it for the newest models.
  • Non-commercial restriction on trial keys — a licensing limit, not a technical one, but it will bite in production.

Key rotation runbook

  • Second key: additional trial keys can be created from the API Keys dashboard page (unconfirmed whether multiple trial keys share the same 1,000/month pool — assume they do; verify on first signup).
  • Revoke: delete/regenerate from dashboard.cohere.com/api-keys (exact button label unconfirmed).
  • ModelHubby integration: export COHERE_API_KEY_2="..." — the key pool picks it up automatically; state machine cools down rate-limited keys per-key.

ModelHubby usage

pnpm mh show cohere
pnpm mh probe --live --provider cohere --apply