Bedrock models with tool use / function calling
These are the Amazon Bedrock models that can call functions — the prerequisite for anything agentic. A model without tool use can answer from what is in the prompt, but it cannot look something up, take an action, or drive a loop.
33 of the models on Bedrock qualify today, from $0.035 per 1M input tokens. Rebuilt daily from the AWS catalogue, so this list does not go stale.
What teams get wrong about this
Tool use is necessary but not sufficient for an agent. The second thing to check is prompt caching: an agent resends its system prompt and tool definitions on every single step, so without caching you re-pay full price for the same prefix dozens of times per task. The caching column below is usually the one that decides the bill.
All 33 models, cheapest first
| Model | Input /1M | Caching |
|---|---|---|
| Nova Micro amazon.nova-micro-v1:0 | $0.035 | yes |
| Nova Lite amazon.nova-lite-v1:0 | $0.060 | yes |
| GPT-5.6 Luna openai.gpt-5.6-luna | $0.200 | yes |
| Llama 3.1 8B Instruct meta.llama3-1-8b-instruct-v1:0 | $0.220 | no |
| Claude 3 Haiku anthropic.claude-3-haiku-20240307-v1:0 | $0.250 | no |
| Llama 3.1 70B Instruct meta.llama3-1-70b-instruct-v1:0 | $0.720 | no |
| Llama 3.3 70B Instruct meta.llama3-3-70b-instruct-v1:0 | $0.720 | no |
| Nova Pro amazon.nova-pro-v1:0 | $0.800 | yes |
| Claude Haiku 4.5 anthropic.claude-haiku-4-5-20251001-v1:0 | $1.00 | yes |
| Mistral Small (24.02) mistral.mistral-small-2402-v1:0 | $1.00 | no |
| Mistral Large (24.07) mistral.mistral-large-2407-v1:0 | $2.00 | no |
| GPT-5.6 Terra openai.gpt-5.6-terra | $2.00 | yes |
| Grok 4.6 xai.grok-4.6 | $2.00 | yes |
| Nova Premier amazon.nova-premier-v1:0 | $2.50 | yes |
| Claude 3 Sonnet anthropic.claude-3-sonnet-20240229-v1:0 | $3.00 | no |
| Claude Sonnet 4.5 anthropic.claude-sonnet-4-5-20250929-v1:0 | $3.00 | yes |
| Claude Sonnet 4.6 anthropic.claude-sonnet-4-6 | $3.00 | yes |
| Claude Sonnet 5 anthropic.claude-sonnet-5 | $3.00 | yes |
| Mistral Large (24.02) mistral.mistral-large-2402-v1:0 | $4.00 | no |
| GPT-5.6 Sol openai.gpt-5.6-sol | $4.00 | yes |
| Claude Opus 4.5 anthropic.claude-opus-4-5-20251101-v1:0 | $5.00 | yes |
| Claude Opus 4.6 anthropic.claude-opus-4-6-v1 | $5.00 | yes |
| Claude Opus 4.7 anthropic.claude-opus-4-7 | $5.00 | yes |
| Claude Opus 4.8 anthropic.claude-opus-4-8 | $5.00 | yes |
| Claude Opus 5 anthropic.claude-opus-5 | $5.00 | yes |
| Claude Fable 5 anthropic.claude-fable-5 | $10.00 | yes |
| Jamba 1.5 Large ai21.jamba-1-5-large-v1:0 | — | no |
| Jamba 1.5 Mini ai21.jamba-1-5-mini-v1:0 | — | no |
| Claude 3.5 Sonnet anthropic.claude-3-5-sonnet-20240620-v1:0 | — | no |
| Claude 3.5 Sonnet v2 anthropic.claude-3-5-sonnet-20241022-v2:0 | — | no |
| Claude 3.7 Sonnet anthropic.claude-3-7-sonnet-20250219-v1:0 | — | no |
| Claude Opus 4.1 anthropic.claude-opus-4-1-20250805-v1:0 | — | yes |
| Claude Sonnet 4 anthropic.claude-sonnet-4-20250514-v1:0 | — | yes |
Capability flags come from the AWS Bedrock model catalogue; prices from the Price List API. Region counts include cross-region inference profiles. Methodology.
Questions
Do all Bedrock models support function calling?
No. Tool use is a per-model capability that Bedrock reports in the model catalogue, and a large share of the models on Bedrock — including most image, video and embedding models, plus several older text models — do not have it. The list on this page is generated from that catalogue, so it reflects what the API will actually accept.
What is the cheapest Bedrock model with tool use?
The table on this page is sorted cheapest-first by input token price, so the top row is the answer. Be careful reading it as the cheapest option overall though: for an agent that resends a long prompt on every step, a dearer model with prompt caching frequently costs less per completed task than a cheap one without it.
Is tool use the same as an agent?
No. Tool use is the model capability; an agent is the loop around it that decides which tool to call, executes it and feeds the result back. You can write that loop yourself against the Converse API, or run it on Amazon Bedrock AgentCore, which is billed separately from tokens.