- Deprecated Models
- 0
- Legacy Models
- 2
- Providers
- 1
- Latest Deprecation
- —
Showing of 2 models
Deprecation history by provider
Why AI models get deprecated — and how to stay ahead
Every AI model has a lifecycle. It launches, becomes widely adopted, and is eventually superseded by a faster, cheaper or more capable successor. Providers can only support so many models at once, so older ones are marked legacy, scheduled for retirement, and finally deprecated. Tracking this lifecycle matters: an app pinned to a discontinued model can break without warning, or quietly cost more than a modern equivalent.
Common reasons models are retired
- A newer generation delivers better quality at a similar or lower price.
- Serving costs make older architectures uneconomical to keep online.
- Safety, licensing or data changes force an older model offline.
- Providers consolidate overlapping models into a single flagship family.
How to migrate without surprises
- Check the Replacement column for the provider's recommended successor.
- Compare input, output and context limits before switching.
- Re-test prompts — newer models often need lighter or different instructions.
- Watch retirement dates so you migrate before the shutdown, not after.
The table above is generated from our live model database and currently covers 0 deprecated and 2 legacy models across 1 provider. Use the provider links below to see a focused deprecation history for a single vendor.
Frequently asked questions
What does "deprecated" mean?
A deprecated AI model is one the provider has officially phased out. It may still respond for a limited grace period, but it no longer receives updates and is scheduled to stop working. Providers usually recommend a newer replacement model.
What does "legacy" mean?
A legacy model is an older generation that remains available but is no longer the recommended option. Legacy models often still run reliably, yet they typically trail newer models on quality, speed, context length or price.
What does "retiring soon" mean?
Retiring soon (scheduled for retirement) means the provider has announced a future date after which the model will be shut down. You can still use it today, but you should plan a migration before the retirement date.
Can I still use these models?
Sometimes. Legacy and retiring-soon models usually still work until their shutdown date, while fully deprecated models may already be unavailable via the API. Always check the provider’s official documentation for the exact status.
What should I use instead?
Where a provider names a successor, we show it in the Replacement column so you can jump straight to the recommended model. If no replacement is listed, compare active models by price and capability to pick the best fit.
How often is this list updated?
The list is generated from our live model database, so counts, dates and replacements update automatically as providers announce deprecations and retirements.