Amazon Bedrock models: every provider in the catalog, and how to choose.
On 10 October 2026 Amazon Bedrock listed models from 19 providers behind one API, including Anthropic Claude, Amazon Nova, OpenAI GPT and gpt-oss, Meta Llama, Mistral, Google Gemma, DeepSeek, Qwen and xAI Grok. Most models are enabled by default. Below: the full provider list, how to pick a model, how access works and how pricing is shaped.
- Bedrock's catalog on 10 October 2026 covered 19 providers: AI21 Labs, Amazon, Anthropic, Cohere, DeepSeek, Google, Luma AI, Meta, MiniMax, Mistral AI, Moonshot AI, NVIDIA, OpenAI, Qwen, Stability AI, TwelveLabs, Writer, xAI and Z.AI. OpenAI models include GPT-6, GPT-5.x and the open-weight gpt-oss models.
- There is no single best model. Pick by task: small, cheap models for high-volume easy calls; mid-tier models as the production default; the top tier for the hardest steps; embedding models for search and RAG; image, video and speech models for media. Most production systems use several, and the Converse API makes switching close to a one-line change.
- Access to most serverless models is on by default if your IAM role has the AWS Marketplace permissions; Anthropic models need a one-time use-case form. Your prompts are not shared with model providers or used to train base models. A partner can file POC funding for a Bedrock build when the workload qualifies and AWS approves.
IEvery model provider on Amazon Bedrock
The provider list and example models, as AWS listed them on 10 October 2026. Availability differs by Region, and the list changes often.
Use the table to see who is on Bedrock and what kind of model each provider offers. For model IDs, context windows, Regions and prices, open the model's card from the models at a glance page and check the pricing page.
| Provider | Example models | Type |
|---|---|---|
| Anthropic | Claude Opus, Sonnet and Haiku (5.x and 4.x), Claude Fable; Claude Mythos is gated | Text, vision, tools |
| Amazon | Nova 2 Lite, Nova Pro, Lite and Micro; Nova 2 Sonic; Nova Canvas; Nova Reel; Nova Multimodal Embeddings; Titan Text Embeddings V2 | Text, speech, image, video, embeddings |
| OpenAI | GPT-6 and GPT-6.1 models, GPT-5.6, GPT-5.5, GPT-5.4; open-weight gpt-oss-120b and gpt-oss-20b; gpt-oss-safeguard | Text, reasoning |
| Meta | Llama 4 Maverick and Scout; Llama 3.3 70B; Llama 3.2 and 3.1 models | Text, vision |
| Mistral AI | Mistral Large 3, Pixtral Large, Devstral 2, Magistral Small, Ministral, Voxtral | Text, vision, code, speech |
| Gemma 4 and Gemma 3 models | Text (open-weight) | |
| DeepSeek | DeepSeek V3.2, V3.1, R1 | Text, reasoning |
| Qwen | Qwen3 235B, Qwen3 32B, Qwen3 Coder models, Qwen3 VL | Text, code, vision |
| xAI | Grok 4.7, 4.6, 4.3 | Text |
| Moonshot AI | Kimi K3, K2.5, K2 Thinking | Text, reasoning |
| MiniMax | MiniMax M2.5, M2.1, M2 | Text |
| Z.AI | GLM 5.3, GLM 5, GLM 4.7, GLM 4.7 Flash | Text |
| NVIDIA | Nemotron 3 Super 120B, Nemotron Nano models | Text, vision |
| Cohere | Command R and R+; Embed v4, English and Multilingual; Rerank 3.5 | Text, embeddings, rerank |
| AI21 Labs | Jamba 1.5 Large and Mini | Text, long context |
| Writer | Palmyra X5, X4, Vision 7B | Text, vision |
| Stability AI | Stable Image Ultra and Core, Stable Diffusion 3.5 Large, image editing tools | Image |
| Luma AI | Ray2 | Video |
| TwelveLabs | Pegasus (video understanding), Marengo Embed | Video, embeddings |
IIHow to choose a model, and why you usually pick several
The catalog is a cost and capability curve. Pick the cheapest model that clears your quality bar for each job.
Four things separate models: capability on your task, price per token (or per image or second), latency, and modality. Context window and Region availability can rule a model out. Run two or three candidates through Bedrock model evaluation on your own prompts, then route traffic by difficulty; Intelligent Prompt Routing can do that per request for supported model pairs.
| Tier | Examples | Use it for | Avoid it for |
|---|---|---|---|
| Small and cheap | Nova Micro and Lite, Claude Haiku, small Llama, Mistral and Gemma models, gpt-oss-20b | Classification, extraction, routing, simple drafts at high volume | Multi-step reasoning |
| Production default | Claude Sonnet, Nova Pro, Mistral Large, larger Llama and Qwen models, GPT-5.x | Chat, coding, agents, most product features | Trivial calls (overpaying) |
| Top tier | Claude Opus, GPT-6 models | The hardest reasoning and code, used on purpose | High-volume or latency-sensitive paths |
| Reasoning-tuned | DeepSeek R1, Kimi K2 Thinking, extended thinking modes | Math, complex code, multi-step analysis | Simple, high-throughput tasks |
| Embeddings and rerank | Titan Text Embeddings V2, Nova Multimodal Embeddings, Cohere Embed and Rerank | Search, RAG, clustering | Producing readable text |
| Media | Nova Canvas, Stable Image, Nova Reel, Luma Ray2, Nova 2 Sonic | Images, video, speech | Text or structured data |
IIIHow model access works
AWS changed this. Most models no longer need an access request.
- Default access: Per the model access page, access to many models is enabled by default in every commercial Region where the model runs, if your IAM role has the AWS Marketplace permissions (Subscribe, Unsubscribe, ViewSubscriptions). The first call to a third-party model starts the subscription in the background.
- Anthropic needs a one-time form: Before the first Claude call, submit use-case details once per account (or once at the organization's management account) in the console or with the PutUseCaseForModelAccess API.
- Some models skip Marketplace: Models from Amazon, DeepSeek, Google, Meta, MiniMax, Mistral AI, Moonshot AI, NVIDIA, Qwen, xAI and Z.AI are not sold through AWS Marketplace, and neither are OpenAI's gpt-oss models.
- Block what you do not want: Because access is on by default, use IAM or SCP deny policies on the Bedrock invoke actions for models your company has not approved.
- Call it by model ID: Use the Converse API for chat models so switching models is close to a one-line change, or InvokeModel for model-specific features. Cross-Region inference profiles spread calls across Regions in a geography.
Step-by-step setup is on our Bedrock setup page.
IVHow Bedrock models are priced
The model you pick sets the rate. How you call it sets the discount.
- On-demand: Pay per input and output token (per image or second for media models). Prices differ widely across the catalog.
- Batch inference: For non-interactive jobs. The pricing page lists select models at 50% lower than on-demand.
- Prompt caching: Repeated prompt prefixes bill at a lower cache-read rate on supported models. See prompt caching.
- Provisioned Throughput: Reserved capacity billed hourly per model unit, with no-commitment, 1-month or 6-month terms. Required to run custom models. See Provisioned Throughput.
See Amazon Bedrock pricing for the full breakdown.
VWhat happens to your data
The same rules apply across the catalog.
The Bedrock security page says your data is not shared with model providers and is not used to improve the base models, that data is encrypted in transit and at rest, and that you can reach Bedrock privately from your VPC with AWS PrivateLink. Fine-tuning works on a private copy of the model. See Bedrock security and compliance.
VIWho can file AWS funding for a Bedrock build
There are two kinds of AWS money for a Bedrock build. A partner files one. You apply for the other yourself.
Partner-filed funding. An AWS consulting partner can file POC (proof of concept) funding to test a Bedrock workload, whether or not you run on AWS today. If you already run on AWS and are adding new AI work, the partner can file incremental workload (IW) funding instead. For startups this is often $50K to $150K. Depends on the workload; AWS approves. The work has to be real, and the partner has to be the one doing it.
AWS Activate, which you apply for yourself. Activate Founders gives $1,000 to start, up to $5,000, with no investor needed. Activate Portfolio gives up to $200,000 if your VC, accelerator or angel group is an Activate Provider and gives you its Org ID. A consulting partner cannot apply for Activate on your behalf. AWS made Activate credits usable for third-party models on Bedrock in April 2024.
For most startups without a VC, the partner route is the larger of the two. CloudRoute matches you with one vetted AWS partner for free. Partners pay CloudRoute a monthly membership plus 10% of AWS-approved funding.
| Funding | Who files it | Amount | When it fits |
|---|---|---|---|
| Proof of concept (POC) funding | Your matched AWS partner | $50K to $150K for startups. Depends on the workload; AWS approves. | You want to test something, whether or not you are on AWS today. |
| Incremental workload (IW) funding | Your matched AWS partner | $50K to $150K for startups. Depends on the workload; AWS approves. | You are already on AWS and are launching new work: a new AI feature, a new customer or a new contract. |
| AWS Activate Founders | Founder applies | $1,000 to start, up to $5,000 | Self-funded startups with no investor or accelerator behind them. |
| AWS Activate Portfolio | Founder applies | up to $200,000 | Startups backed by a VC firm, accelerator or angel investor that is an AWS Activate Provider. |
| AWS Credits for AI Startups | Invite only (AWS account manager) | $200,000 or more, by invitation | AI startups ready to scale after Activate Portfolio, invited by AWS. |
How a match works (example)
- You tell us about the work
You fill in a short form: what you run today, what you want to move, test or build, your stage and your region.
- We pick one partner
We read your answers and pick one vetted AWS consulting partner that fits the work, the stack and the region. We reply within 24 hours; your matched partner reaches out within 5 business days.
- The partner checks POC funding
Your matched partner scopes the test with you. Then it checks whether the test qualifies for POC (proof of concept) funding. A POC can fit whether or not you run on AWS today.
- The partner files, AWS decides
If the work qualifies, the partner files the funding request with AWS. AWS reviews it and decides. The amount depends on the workload.
- You pay nothing
CloudRoute is free for startups. Partners pay CloudRoute a monthly membership plus 10% of AWS-approved funding.
AWS Activate credits are separate, and you apply for them yourself. Use Activate Founders if no investor backs you. Use Activate Portfolio with the Org ID or invite from your VC, accelerator, incubator or angel group if it is an Activate Provider. A consulting partner cannot apply for Activate on your behalf.
Common questions
Which models are available on Amazon Bedrock in 2026?
Are OpenAI models available on Amazon Bedrock?
What is the best model on Amazon Bedrock?
Do I need to request access to Bedrock models?
Can I switch between Bedrock models without rewriting code?
Do Bedrock models keep my data private?
Can AWS funding pay for a Bedrock build?
Get a partner who can build it and file the funding
Tell us what you are building on Bedrock. We match you with one vetted AWS partner. If the work qualifies, the partner files POC funding for a test, or incremental workload (IW) funding for new work on AWS: often $50K to $150K for startups. Depends on the workload; AWS approves. CloudRoute is free for startups.