Cohere · api
fine-tuning deprecation: shutdown date & replacement
Shut down September 15, 2025
- Provider
- Cohere
- Name
- Fine-tuning (command-light, command, command-r, classify, rerank)
- Type
- api
- Announced
- —
- Shutdown
- Replacement
- —
- Source
- Official Cohere notice
Fine-tuning retired effective September 15, 2025 in dashboard and API; previously fine-tuned models no longer accessible.
How to migrate off fine-tuning
- Find every place that references it:
grep -rn --exclude-dir={node_modules,.git,dist} "fine-tuning" . - Follow the provider’s migration notes in the official notice.
- Re-run your evals or a sample of real prompts — output style, token usage and pricing can change between APIs.
- Check config outside code too: environment variables, feature flags, prompt-management tools, and fine-tuned model references.
- Deploy before September 15, 2025, and add the CI scanner so it cannot come back.
Same model on other platforms
- OpenAI — January 6, 2027
Questions
Is fine-tuning deprecated?
Yes. Cohere has deprecated fine-tuning. It was shut down on September 15, 2025.
When does fine-tuning shut down?
September 15, 2025, according to Cohere's official deprecation notice.
What should I use instead of fine-tuning?
Cohere has not named a specific replacement. Check the official source for current models.
What happens to requests to fine-tuning after the shutdown?
Requests using the retired model ID fail with an error, so any code still calling it breaks. Update and deploy before the shutdown date.