Coding Agents

AI model ID reference: Claude, GPT, Gemini and more

Q: What is a model ID?

It is the exact string you pass as the model parameter in an API call, such as claude-opus-5-5, gpt-5.5 or gemini-3.8-flash. It is not the marketing name: “Claude Opus 5.5” or “Gemini 3.8 Flash” will be rejected. IDs are case-sensitive, and cloud platforms such as Amazon Bedrock often use a different form of the same model’s ID.

Q: How do I find the models my API key can use?

Call the provider’s list-models endpoint: GET https://api.anthropic.com/v1/models for Claude, GET https://api.openai.com/v1/models for OpenAI, and GET https://generativelanguage.googleapis.com/v1beta/models for Gemini. Each returns the IDs your key can call, and the API key checker makes the same request from your browser. This reference adds what those endpoints don’t tell you: aliases, cloud IDs and retirement dates.

Copy the exact API model ID for Claude, GPT, Gemini, Grok and DeepSeek models, with aliases, pinned snapshots, Bedrock and Vertex AI forms and retirement dates, all checked against each provider’s docs. Free, and it runs in your browser.

Search by model name or ID

94 models. Click an ID to copy it.

Claude Fable 5.1Anthropic
CurrentRecommended
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 1 Sep 2027.
Pricing and specs Source
Claude Opus 5.5Anthropic
CurrentRecommended
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 22 Sep 2027.
Pricing and specs Source
Claude Sonnet 5.5Anthropic
CurrentRecommended
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 28 Sep 2027.
Pricing and specs Source
Claude Haiku 5.5Anthropic
CurrentRecommended
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 7 Oct 2027.
Pricing and specs Source
Claude Mythos 5.1Anthropic
CurrentRestricted access
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 1 Sep 2027.
Only for organisations verified through Anthropic’s verification programmes. Same specifications as Claude Fable 5.1.
Source
GPT-6 AstraOpenAI
CurrentRecommended
1.05M context · 128K max output
API ID
- Pinned
Source
GPT-6.1 SolOpenAI
CurrentRecommended
1.05M context · 128K max output
API ID
- Pinned
Pricing and specs Source
GPT-6 LunaOpenAI
CurrentRecommended
1.05M context · 128K max output
API ID
- Pinned
Pricing and specs Source
GPT-5.6 SolOpenAI
Current
1.05M context · 128K max output
API ID
- Pinned
- Alias
  Routes requests to GPT-5.6 Sol.
Source
GPT-5.6 TerraOpenAI
Current
1.05M context · 128K max output
API ID
- Pinned
Source
GPT-5.6 LunaOpenAI
Current
1.05M context · 128K max output
API ID
- Pinned
Source
GPT-5.6 CyberOpenAI
CurrentRestricted access
400K context · 128K max output
API ID
- Alias
  An alias for OpenAI’s most advanced cybersecurity models.
Needs separate approval: OpenAI asks defenders to apply through its Daybreak programme.
Source
GPT-5.5OpenAI
Current
1.05M context · 128K max output
API ID
- Alias
- Pinned
Source
GPT-5.5 ProOpenAI
Current
1.05M context · 128K max output
API ID
- Alias
- Pinned
Source
GPT-5.4OpenAI
Current
1.05M context · 128K max output
API ID
- Alias
- Pinned
Source
GPT-5.4 ProOpenAI
Current
1.05M context · 128K max output
API ID
- Alias
- Pinned
Source
GPT-5.4 miniOpenAI
Current
400K context · 128K max output
API ID
- Alias
- Pinned
Source
GPT-4oOpenAI
Current
128K context · 16,384 max output
API ID
- Alias
  Points to gpt-4o-2024-08-06.
- Pinned
- Pinned
- Pinned
  Deprecated: shuts down on 23 October 2026 (replacement gpt-5.6-sol).
Source
GPT-4o miniOpenAI
Current
128K context · 16,384 max output
API ID
- Alias
- Pinned
Source
Chat LatestOpenAI
Current
400K context · 128K max output
API ID
- Alias
  Points to the latest Instant model used in ChatGPT; the snapshot behind it is updated regularly.
OpenAI lists it under ChatGPT models, not recommended for API use, and recommends GPT-6 Astra for production.
Source
Gemini 3.8 FlashGoogle
CurrentRecommended
1,048,576 context · 65,536 max output
API ID
- Stable
Pricing and specs Source
Gemini 3.5 Flash-LiteGoogle
CurrentRecommended
1,048,576 context · 65,536 max output
API ID
- Stable
Pricing and specs Source
Gemini 3.1 Pro PreviewGoogle
CurrentPreview
1,048,576 context · 65,536 max output
API ID
- Preview
- Preview
Source
Grok 4.7xAI
CurrentRecommended
500K context
API ID
- Alias
Pricing and specs Source
Grok 4.6xAI
Current
500K context
API ID
- Alias
Source
Grok 4.5xAI
Current
500K context
API ID
- Alias
- Alias
- Alias
Source
Grok 4.3xAI
Current
1M context
API ID
- Alias
- Alias
Source
Grok 4.20 (reasoning)xAI
Current
1M context
API ID
- Pinned
- Alias
- Alias
- Alias
- Alias
xAI also accepts several older beta and experimental names for this model.
Source
Grok 4.20 (non-reasoning)xAI
Current
1M context
API ID
- Pinned
- Alias
- Alias
xAI also accepts several older beta and experimental names for this model.
Source
Grok 4.20 Multi-AgentxAI
Current
1M context
API ID
- Pinned
- Alias
- Alias
xAI also accepts several older beta and experimental names for this model.
Source
Grok Build 0.1xAI
Current
256K context
API ID
- Alias
- Redirect
- Redirect
- Redirect
The grok-code-fast names now resolve to Grok Build 0.1.
Source
DeepSeek V4.1 FlashDeepSeek
CurrentRecommended
1M context · 384K max output
API ID
- Alias
- Redirect
- Redirect
The older deepseek-v4-flash names are still accepted, but their models are retired and requests are served by V4.1 Flash at the Flash price.
Pricing and specs Source
DeepSeek V4 ProDeepSeek
CurrentRecommended
1M context · 384K max output
API ID
- Alias
DeepSeek has said it will keep serving V4 Pro after 14 September 2026 and will give notice of any change.
Pricing and specs Source
Claude Fable 5Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 9 Jun 2027. Replacement: claude-fable-5-1
Source
Claude Mythos 5Anthropic
LegacyRestricted access
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 9 Jun 2027. Replacement: claude-mythos-5-1
Only for organisations verified through Anthropic’s verification programmes. Anthropic publishes a migration guide to Claude Mythos 5.1.
Source
Claude Opus 5Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 24 Jul 2027. Replacement: claude-opus-5-5
Source
Claude Sonnet 5Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 30 Jun 2027. Replacement: claude-sonnet-5-5
Source
Claude Opus 4.8Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 28 May 2027. Replacement: claude-opus-5-5
Source
Claude Opus 4.7Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Amazon Bedrock
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 16 Apr 2027. Replacement: claude-opus-5-5
Source
Claude Opus 4.6Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-opus-4-6-v1
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 5 Feb 2027. Replacement: claude-opus-5-5
Source
Claude Sonnet 4.6Anthropic
Legacy
1M context · 128K max output
API ID
- Pinned
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-sonnet-4-6
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 17 Feb 2027. Replacement: claude-sonnet-5-5
Source
Claude Opus 4.5Anthropic
Legacy
200K context · 64K max output
API ID
- Pinned
- Alias
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-opus-4-5-20251101-v1:0
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 24 Nov 2026. Replacement: claude-opus-5-5
Source
Claude Haiku 4.5Anthropic
Legacy
200K context · 64K max output
API ID
- Pinned
- Alias
Cloud platforms
Amazon Bedrock
Bedrock InvokeModelInference profiles for base ID anthropic.claude-haiku-4-5-20251001-v1:0
Google Cloud
Microsoft FoundryDefault deployment name
Retires no sooner than 15 Oct 2026. Replacement: claude-haiku-5-5
Source
GPT-6 SolOpenAI
Legacy
1.05M context · 128K max output
API ID
- Pinned
Replacement: gpt-6.1-sol
OpenAI points to GPT-6.1 Sol as the newer Sol model.
Source
GPT-5.2OpenAI
Legacy
400K context · 128K max output
API ID
- Alias
- Pinned
Replacement: gpt-6-astra
OpenAI calls it its previous flagship and recommends GPT-6 Astra.
Source
GPT-5.2 ProOpenAI
Legacy
400K context · 128K max output
API ID
- Alias
- Pinned
Replacement: gpt-5.5-pro
OpenAI recommends GPT-5.5 Pro as the latest pro model.
Source
GPT-4.1OpenAI
Legacy
1,047,576 context · 32,768 max output
API ID
- Alias
- Pinned
A non-reasoning model. OpenAI’s page recommends starting with a GPT-5 model for more complex tasks.
Source
GPT-4.1 miniOpenAI
Legacy
1,047,576 context · 32,768 max output
API ID
- Alias
- Pinned
A non-reasoning model. OpenAI’s page recommends starting with a GPT-5 model for more complex tasks.
Source
Gemini 3.6 FlashGoogle
Legacy
1,048,576 context · 65,536 max output
API ID
- Stable
Replacement: gemini-3.8-flash
Google calls it its previous-generation Flash model.
Source
Gemini 3.7 FlashGoogle
Legacy
API ID
- Redirect
Requests to gemini-3.7-flash are automatically routed to gemini-3.8-flash.
Source
Gemini 3.5 FlashGoogle
Legacy
API ID
- Redirect
Requests to gemini-3.5-flash are automatically routed to gemini-3.6-flash, although Google’s quickstart still uses this code in its examples.
Source
Gemini 3 Flash PreviewGoogle
LegacyPreview
1,048,576 context · 65,536 max output
API ID
- Preview
Replacement: gemini-3.6-flash
No shutdown date announced; Google recommends gemini-3.6-flash.
Source
Gemini 2.5 ProGoogle
Legacy
1,048,576 context · 65,536 max output
API ID
- Stable
Not deprecated, but Google limits the 2.5 models to accounts that have used them before. New projects should use Gemini 3.8 Flash or 3.5 Flash-Lite.
Source
Gemini 2.5 FlashGoogle
Legacy
1,048,576 context · 65,536 max output
API ID
- Stable
Not deprecated, but Google limits the 2.5 models to accounts that have used them before. New projects should use Gemini 3.8 Flash or 3.5 Flash-Lite.
Source
Gemini 2.5 Flash-LiteGoogle
Legacy
1,048,576 context · 65,536 max output
API ID
- Stable
Not deprecated, but Google limits the 2.5 models to accounts that have used them before. New projects should use Gemini 3.8 Flash or 3.5 Flash-Lite.
Source
Claude Sonnet 4.5Anthropic
Deprecated
200K context · 64K max output
API ID
- Pinned
- Alias
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-sonnet-4-5-20250929-v1:0
Google Cloud
Microsoft FoundryDefault deployment name
Retires 30 Nov 2026. Replacement: claude-sonnet-5-5
Source
Claude Mythos PreviewAnthropic
DeprecatedPreviewRestricted access
API ID
- Preview
Cloud platforms
Amazon Bedrock
Invitation only, through Anthropic’s Project Glasswing. Deprecated on 9 June 2026; Anthropic has not announced a retirement date yet.
Source
GPT-5.1OpenAI
Deprecated
400K context · 128K max output
API ID
- Alias
- Pinned
Retires 1 Apr 2027. Replacement: gpt-6-sol
Source
GPT-5.3-CodexOpenAI
Deprecated
400K context · 128K max output
API ID
- Pinned
Retires 1 Apr 2027. Replacement: gpt-6-sol
Source
GPT-5.4 nanoOpenAI
Deprecated
400K context · 128K max output
API ID
- Alias
- Pinned
Retires 1 Apr 2027. Replacement: gpt-6-luna
Source
GPT-5OpenAI
Deprecated
400K context · 128K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-sol
OpenAI’s deprecations page schedules gpt-5-2025-08-07, the only snapshot behind the gpt-5 alias, for shutdown.
Source
GPT-5 miniOpenAI
Deprecated
400K context · 128K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-terra
OpenAI’s deprecations page schedules gpt-5-mini-2025-08-07, the only snapshot behind the alias, for shutdown.
Source
GPT-5 nanoOpenAI
Deprecated
400K context · 128K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-luna
OpenAI’s deprecations page schedules gpt-5-nano-2025-08-07, the only snapshot behind the alias, for shutdown.
Source
GPT-5 ProOpenAI
Deprecated
400K context · 272K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-sol
OpenAI’s suggested replacement is gpt-5.6-sol with reasoning.mode set to pro.
Source
o3OpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-sol
Source
o3-proOpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 11 Dec 2026. Replacement: gpt-5.6-sol
OpenAI’s suggested replacement is gpt-5.6-sol with reasoning.mode set to pro.
Source
GPT-4.1 nanoOpenAI
Deprecated
1,047,576 context · 32,768 max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-luna
Source
o4-miniOpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-terra
Source
o3-miniOpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-sol
Source
o1OpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-sol
Source
o1-proOpenAI
Deprecated
200K context · 100K max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-sol
OpenAI’s suggested replacement is gpt-5.6-sol with reasoning.mode set to pro.
Source
GPT-4 TurboOpenAI
Deprecated
128K context · 4,096 max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-sol
Source
GPT-4OpenAI
Deprecated
8,192 context · 8,192 max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-sol
Source
GPT-3.5 TurboOpenAI
Deprecated
16,385 context · 4,096 max output
API ID
- Alias
- Pinned
Retires 23 Oct 2026. Replacement: gpt-5.6-terra
gpt-3.5-turbo-1106 and gpt-3.5-turbo-instruct were shut down on 28 September 2026.
Source
Gemini 3.1 Flash-LiteGoogle
Deprecated
1,048,576 context · 65,536 max output
API ID
- Stable
Shuts down 7 May 2027 at the earliest. Replacement: gemini-3.5-flash-lite
Source
Claude Opus 4.1Anthropic
Retired
API ID
- Pinned
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-opus-4-1-20250805-v1:0
Google Cloud
Retired 5 Aug 2026. Replacement: claude-opus-4-8
Retired on the Claude API. Amazon Bedrock and Google Cloud set their own dates and still list it as deprecated.
Source
Claude Sonnet 4Anthropic
Retired
API ID
- Pinned
Cloud platforms
Bedrock InvokeModelInference profiles for base ID anthropic.claude-sonnet-4-20250514-v1:0
Google Cloud
Retired 15 Jun 2026. Replacement: claude-sonnet-5-5
Retired on the Claude API. Amazon Bedrock and Google Cloud set their own dates and still list it as deprecated.
Source
Claude Opus 4Anthropic
Retired
API ID
- Pinned
Cloud platforms
Google Cloud
Retired 15 Jun 2026. Replacement: claude-opus-4-8
Retired on the Claude API. Google Cloud sets its own dates and still lists it as deprecated.
Source
Claude Haiku 3Anthropic
Retired
API ID
- Pinned
Retired 20 Apr 2026. Replacement: claude-haiku-4-5-20251001
Source
Claude Sonnet 3.7Anthropic
Retired
API ID
- Pinned
Retired 19 Feb 2026. Replacement: claude-sonnet-5-5
Source
Claude Haiku 3.5Anthropic
Retired
API ID
- Pinned
Cloud platforms
Google Cloud
Retired 19 Feb 2026. Replacement: claude-haiku-4-5-20251001
Retired on the Claude API. Google Cloud sets its own dates and still lists it as deprecated.
Source
Claude Opus 3Anthropic
Retired
API ID
- Pinned
Retired 5 Jan 2026. Replacement: claude-opus-4-8
Source
GPT-5.2 Chat and GPT-5.3 ChatOpenAI
Retired
API ID
- Alias
- Alias
Retired 10 Aug 2026. Replacement: gpt-5.6-sol
Source
GPT-5.2-CodexOpenAI
Retired
API ID
- Pinned
Retired 23 Jul 2026. Replacement: gpt-5.6-sol
Source
GPT-5.1-Codex family and GPT-5-CodexOpenAI
Retired
API ID
- Pinned
- Pinned
- Pinned
  OpenAI’s suggested replacement is gpt-5.6-terra.
- Pinned
Retired 23 Jul 2026. Replacement: gpt-5.6-sol
Source
GPT-5 Chat and GPT-5.1 ChatOpenAI
Retired
API ID
- Alias
- Alias
Retired 23 Jul 2026. Replacement: gpt-5.6-sol
Source
o3 and o4-mini deep researchOpenAI
Retired
API ID
- Alias
- Pinned
- Alias
- Pinned
Retired 23 Jul 2026. Replacement: gpt-5.6-sol
Source
Gemini 2.0 FlashGoogle
Retired
API ID
- Stable
- Stable
Retired 1 Jun 2026. Replacement: gemini-3.6-flash
Source
Gemini 2.0 Flash-LiteGoogle
Retired
API ID
- Stable
- Stable
Retired 1 Jun 2026. Replacement: gemini-3.1-flash-lite
Source
Gemini 3.1 Flash-Lite PreviewGoogle
RetiredPreview
API ID
- Preview
Retired 25 May 2026. Replacement: gemini-3.1-flash-lite
Source
Gemini 2.5 Flash-Lite PreviewGoogle
RetiredPreview
API ID
- Preview
Retired 31 Mar 2026. Replacement: gemini-3.1-flash-lite
Source
Gemini 3 Pro PreviewGoogle
RetiredPreview
API ID
- Preview
Retired 9 Mar 2026. Replacement: gemini-3.1-pro-preview
Source
Grok 4.1 Fast, Grok 4 Fast, Grok 4 and Grok 3xAI
Retired
API ID
- Redirect
- Redirect
- Redirect
- Redirect
- Redirect
- Redirect
Retired 15 May 2026. Replacement: grok-4.3
Retired on 15 May 2026. The names still resolve: reasoning names are served by grok-4.3 at low reasoning effort, non-reasoning names (and grok-3) at none, billed at Grok 4.3 prices.
Source
deepseek-chat and deepseek-reasonerDeepSeek
Retired
API ID
- Alias
- Alias
Retired 24 Jul 2026. Replacement: deepseek-flash
With the V4 release DeepSeek announced these legacy names would be discontinued on 24 July 2026; they no longer appear on its models page.
Source

Steps

How to use the model ID reference

Type part of a model name or ID in the search box, such as “opus”, “gpt-5” or “flash”.
Narrow the list by provider or status. Retired models stay listed so you can check an old ID.
Click an ID to copy it. Prefer a Pinned or Stable ID in production code.
For Claude on Amazon Bedrock, Google Cloud or Microsoft Foundry, copy the ID from the Cloud platforms rows.
Open “Show code” to see the SDK call with that ID in place, and check the retirement date before you ship.

Method

How it works

A model ID is the string that goes in the model field of an API request. Providers publish the IDs on their own model pages, but each one uses a different naming scheme, the same model can have several IDs, and old IDs stop working on fixed dates. This reference brings the text models from Anthropic (Claude), OpenAI (GPT), Google (Gemini), xAI (Grok) and DeepSeek (DeepSeek) into one list, with the status of each and the date it retires.

Aliases, snapshots and pinned IDs

Some IDs always mean the same model; others are pointers that move. We label every ID with what its provider says it does:

Pinned: always the same model version. Every Claude ID is pinned, including dateless ones like claude-opus-5-5. OpenAI’s dated snapshots, such as gpt-5.5-2026-04-23, are pinned too.
Stable: a Gemini stable code such as gemini-3.8-flash, which Google says usually doesn’t change and recommends for most production apps.
Alias: a shortcut the provider can move, such as gpt-5.5, claude-haiku-4-5 or grok-4.7. Handy for trying the latest version; risky when you need results to stay the same.
Preview: an early model that can change or be shut down at short notice (Google gives at least two weeks).
Redirect: a retired name the provider now serves with another model, like gemini-3.7-flash. It works today, but you are no longer getting the model you asked for.

Cloud platforms use different IDs

Claude is also sold through Amazon Bedrock, Google Cloud (Vertex AI, now part of its Agent Platform) and Microsoft Foundry. Bedrock adds an anthropic. prefix (anthropic.claude-opus-5-5), and older models called through Bedrock’s InvokeModel API need a cross-region inference profile prefix such as global. or us.. Google Cloud uses the Claude API ID, but writes the date of older snapshots after an @. Foundry calls your deployment, which is named after the model ID unless you rename it. Partner platforms also set their own retirement dates, so a model retired on the Claude API can still run on Bedrock for a while.

Status and retirement dates

We use four statuses. Current means the provider lists the model as current; Legacy means it still works but the provider points to something newer; Deprecated means a shutdown date has been announced; Retired means it has been switched off or redirected. Today the list has 33 current, 22 legacy, 20 deprecated and 19 retired entries. Dates follow each provider’s wording: Anthropic commits to keeping current models “not sooner than” a date, Google lists the earliest possible shutdown, and OpenAI gives a fixed shutdown date. When a model is deprecated, test its replacement well before the date: Claude Sonnet 4.5, for example, retires on 30 Nov 2026 and Anthropic recommends claude-sonnet-5-5.

How this list is kept accurate

Every row comes from the provider’s own model and deprecation pages, never from a third-party list; each model links to its source. We use the LiteLLM model list and our own pricing data only to cross-check. If a provider’s page was ambiguous we left the model out rather than guess, which is why Mistral isn’t here yet, and image, audio and embedding models are left out to keep the list about text models. A script fetches the source pages and flags IDs they list that we don’t; it runs by hand, so the date below is the last time a person checked.

Need prices or limits next? Compare models in the AI model comparison, estimate a bill with the LLM cost calculator, or count your prompt with the token counter.

Examples

Worked examples

Which ID to use in production, by provider

Anthropic (Claude)
Every Claude model ID is a pinned snapshot, including the dateless IDs used from the 4.6 generation on (claude-opus-5-5 never changes). Older models also accept a short alias such as claude-haiku-4-5, which points to the newest dated snapshot of that version: pin the dated ID in production.
List your models: GET https://api.anthropic.com/v1/models
OpenAI (GPT)
A model name such as gpt-5.5 is an alias for its default snapshot. Dated snapshots (gpt-5.5-2026-04-23) lock in one version so behaviour stays the same: pin the snapshot in production. Several newer models list a single snapshot with the same name as the model.
List your models: GET https://api.openai.com/v1/models
Google (Gemini)
Google says most production apps should use a specific stable model code such as gemini-3.8-flash. Preview codes can be shut down with two weeks’ notice, and -latest aliases are hot-swapped to each new release.
List your models: GET https://generativelanguage.googleapis.com/v1beta/models
xAI (Grok)
xAI aliases <modelname> to the latest stable version and <modelname>-latest to the latest version. A dated name such as grok-4.20-0309-reasoning never changes: xAI recommends it for workflows that need consistency.
DeepSeek (DeepSeek)
DeepSeek publishes two model names and updates the model behind each in place, so there is no dated ID to pin. Watch its change log: retired names are sometimes routed to a newer model for a while.

Recommended models at a glance

The models each provider currently recommends, with the ID to copy. Checked 11 Oct 2026.
Model	Amazon Bedrock	Google Cloud	Context
Claude Fable 5.1			1M
Claude Opus 5.5			1M
Claude Sonnet 5.5			1M
Claude Haiku 5.5			1M
GPT-6 Astra	–	–	1.05M
GPT-6.1 Sol	–	–	1.05M
GPT-6 Luna	–	–	1.05M
Gemini 3.8 Flash	–	–	1,048,576
Gemini 3.5 Flash-Lite	–	–	1,048,576
Grok 4.7	–	–	500K
DeepSeek V4.1 Flash	–	–	1M
DeepSeek V4 Pro	–	–	1M

Where the model ID goes in your code

Claude

import anthropic

client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY

message = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=1000,
    messages=[{"role": "user", "content": "Hello, Claude"}],
)
print(message.content)

OpenAI

from openai import OpenAI

client = OpenAI()  # reads OPENAI_API_KEY

response = client.responses.create(
    model="gpt-5.5-2026-04-23",
    input="Write a one-sentence bedtime story about a unicorn.",
)
print(response.output_text)

Gemini

from google import genai

client = genai.Client()  # reads GEMINI_API_KEY

interaction = client.interactions.create(
    model="gemini-3.8-flash",
    input="Explain how AI works in a few words",
)
print(interaction.output_text)

Amazon Bedrock

# Bedrock's Messages-API endpoint (bedrock-mantle)
from anthropic import AnthropicBedrockMantle

client = AnthropicBedrockMantle(aws_region="us-east-1")

message = client.messages.create(
    model="anthropic.claude-opus-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello, Claude"}],
)

Google Cloud

# Vertex AI / Agent Platform
from anthropic import AnthropicVertex

client = AnthropicVertex(project_id="MY_PROJECT_ID", region="global")

message = client.messages.create(
    model="claude-opus-5-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello, Claude"}],
)

Each example follows the provider’s own quickstart. On Google Cloud, the SDK puts the ID in the endpoint URL rather than the request body.

FAQ

Frequently asked questions

What is a model ID?

It is the exact string you pass as the model parameter in an API call, such as claude-opus-5-5, gpt-5.5 or gemini-3.8-flash. It is not the marketing name: “Claude Opus 5.5” or “Gemini 3.8 Flash” will be rejected. IDs are case-sensitive, and cloud platforms such as Amazon Bedrock often use a different form of the same model’s ID.

What is the difference between a model alias and a snapshot?

A snapshot (or pinned ID) always runs the same model version, so your results stay stable. An alias is a shortcut the provider can point at a newer version, so behaviour can change without you editing code. OpenAI’s gpt-5.5 points to the snapshot gpt-5.5-2026-04-23, for example. Use pinned IDs in production and aliases for experiments.

Why does the same Claude model have a different ID on Bedrock and Vertex AI?

Each platform has its own naming scheme. Amazon Bedrock adds an anthropic. prefix (and a -v1:0 suffix on older dated models), while Google Cloud uses the Claude API ID but writes older snapshot dates after an @, as in claude-haiku-4-5@20251001. The model is the same; only the ID changes. This page lists each form so you can copy the right one.

Why do I get a “model not found” error?

The usual causes are a typo or a display name instead of the ID, a model that has been retired, an ID in the wrong platform’s format (a Claude API ID sent to Bedrock, for example), or a model your account or region can’t use yet. Check the ID and its status here, then list the models your key can access with the provider’s models endpoint.

How do I find the models my API key can use?

Call the provider’s list-models endpoint: GET https://api.anthropic.com/v1/models for Claude, GET https://api.openai.com/v1/models for OpenAI, and GET https://generativelanguage.googleapis.com/v1beta/models for Gemini. Each returns the IDs your key can call, and the API key checker makes the same request from your browser. This reference adds what those endpoints don’t tell you: aliases, cloud IDs and retirement dates.

How often is this list checked?

Every row was copied from the provider’s own documentation and last checked on 11 Oct 2026. Providers retire and add models often, so we re-check the source pages with a script that flags any ID they list that we don’t, and review the deprecation pages by hand. Each model links to the page it came from.

Read

Model IDs