Skip to content

Bedrock model cost reference

119 of 119 models · estimated cost in your AWS account, per million tokens

Prices synced from AWS · last updated

Model labs · All labs

Amazon: Nova 2 Lite

text

Nova 2 Lite is a fast, cost-effective reasoning model for everyday workloads that can process text, images, and videos to generate text.

Amazon1M contextUS$0.33/M inputUS$2.75/M output

Amazon: Nova Lite

text

Very low-cost multimodal model that processes text, images and video quickly. A strong default for high-volume workloads.

Amazon300k contextUS$0.06/M inputUS$0.24/M output

Amazon: Nova Micro

text

Text-only model optimized for the lowest latency and cost. Great for classification, routing, extraction and short chat turns.

Amazon128k contextUS$0.035/M inputUS$0.14/M output

Amazon: Nova Pro

text

Highly capable multimodal model with the best balance of accuracy, speed and cost across a wide range of tasks.

Amazon300k contextUS$0.8/M inputUS$3.2/M output

Anthropic: Claude 3 Haiku

text

The fastest, cheapest Claude 3 model.

Anthropic200k contextUS$0.25/M inputUS$1.25/M output

Anthropic: Claude 3.5 Haiku

text

Fast, low-cost Claude for high-volume tasks.

Anthropic200k contextUS$0.8/M inputUS$4/M output

Anthropic: Claude 3.7 Sonnet

text

Hybrid reasoning Claude with extended thinking.

Anthropic200k contextUS$3/M inputUS$15/M output

Anthropic: Claude Fable 5

text

Claude Fable 5 is Anthropic's model for demanding reasoning and long-horizon agentic work, thinking on every request.

Anthropic1M contextUS$11/M inputUS$55/M output

Anthropic: Claude Fable 5.1

text

Claude Fable 5.1 is Anthropic's model for demanding reasoning and long-horizon agentic work, succeeding Claude Fable 5.

Anthropic1M contextUS$11/M inputUS$55/M output

Anthropic: Claude Fable 5.2

text

Claude Fable 5.2 is Anthropic's most capable model for demanding reasoning and long-horizon agentic work, succeeding Claude Fable 5.1.

Anthropic1M contextUS$8.8/M inputUS$44/M output

Anthropic: Claude Haiku 4.5

text

Anthropic's fastest model with near-frontier intelligence. Available through Bring Your Own Key with your AWS account.

Anthropic200k contextUS$1.1/M inputUS$5.5/M output

Anthropic: Claude Haiku 5.5

text

Claude Haiku 5.5 is Anthropic's fastest model, for high-volume, latency-sensitive tasks such as classification, extraction, and routing.

Anthropic1M contextUS$0.11/M inputUS$0.55/M output

Anthropic: Claude Opus 4

text

Claude 4 Opus for long-running agentic and coding work.

Anthropic200k contextUS$15/M inputUS$75/M output

Anthropic: Claude Opus 4.1

text

Opus 4.1 for agentic tasks, coding and reasoning.

Anthropic200k contextUS$15/M inputUS$75/M output

Anthropic: Claude Opus 4.5

text

Claude Opus 4.5 is Anthropic’s frontier reasoning model optimized for complex software engineering, agentic workflows, and long-horizon computer use.

Anthropic200k contextUS$5.5/M inputUS$27.5/M output

Anthropic: Claude Opus 4.6

text

Opus 4.6 is Anthropic’s strongest model for coding and long-running professional tasks.

Anthropic1M contextUS$5.5/M inputUS$27.5/M output

Anthropic: Claude Opus 4.7

text

Opus 4.7 is the next generation of Anthropic's Opus family, built for long-running, asynchronous agents.

Anthropic1M contextUS$5.5/M inputUS$27.5/M output

Anthropic: Claude Opus 4.8

text

Claude Opus 4.8 is Anthropic's most capable generally available model in the Opus family.

Anthropic1M contextUS$5.5/M inputUS$27.5/M output

Anthropic: Claude Opus 5

text

Claude Opus 5 is Anthropic’s flagship model for demanding reasoning, coding, and long-horizon agentic work.

Anthropic1M contextUS$5.5/M inputUS$27.5/M output

Anthropic: Claude Opus 5.5

text

Claude Opus 5.5 is Anthropic's flagship model for demanding reasoning, coding, and long-horizon agentic work, succeeding Claude Opus 5.

Anthropic1M contextUS$4.4/M inputUS$22/M output

Anthropic: Claude Sonnet 4

text

Balanced Claude 4 model for coding and everyday reasoning.

Anthropic200k contextUS$3/M inputUS$15/M output

Anthropic: Claude Sonnet 4.5

text

Anthropic's high-intelligence model for complex agents and coding. Available through Bring Your Own Key with your AWS account.

Anthropic1M contextUS$3.3/M inputUS$16.5/M output

Anthropic: Claude Sonnet 4.6

text

Sonnet 4.6 is Anthropic's most capable Sonnet-class model yet, with frontier performance across coding, agents, and professional work.

Anthropic1M contextUS$3.3/M inputUS$16.5/M output

Anthropic: Claude Sonnet 5

text

Sonnet 5 is Anthropic's most capable Sonnet-class model, with frontier performance across coding, agents, and professional work.

Anthropic1M contextUS$2.2/M inputUS$11/M output

Anthropic: Claude Sonnet 5.5

text

Claude Sonnet 5.5 is Anthropic's best combination of speed and intelligence, succeeding Claude Sonnet 5 for coding, agents, and professional work.

Anthropic1M contextUS$2.2/M inputUS$11/M output

Cohere: Command A

text

Cohere's most performant model for agentic, tool-use and RAG workloads.

cohere256k contextUS$2.5/M inputUS$10/M output

Cohere: Command A Reasoning

text

Command A with reasoning for complex agentic tasks.

cohere256k contextUS$2.5/M inputUS$10/M output

Cohere: Command A Vision

text

Command A with image understanding.

cohere128k contextUS$2.5/M inputUS$10/M output

Cohere: Command R

text

Scalable RAG and tool use at low cost.

cohere128k contextUS$0.15/M inputUS$0.6/M output

Cohere: Command R+

text

Retrieval-augmented generation and multi-step tool use.

cohere128k contextUS$2.5/M inputUS$10/M output

Cohere: Command R7B

text

Small, fast Command model.

cohere128k contextUS$0.037/M inputUS$0.15/M output

DeepSeek: R1

text

Open reasoning model trained with large-scale reinforcement learning. Excels at math, code and multi-step logic.

DeepSeek64k contextUS$1.35/M inputUS$5.4/M output

DeepSeek: V3.2

text

Hybrid reasoning model with sparse attention for long contexts. Strong at coding, tool use and agentic workflows.

DeepSeek164k contextUS$0.62/M inputUS$1.85/M output

Google: Gemini 2.0 Flash

text

Second-generation workhorse Gemini with a 1M-token context.

Google1.0M contextUS$0.1/M inputUS$0.4/M output

Google: Gemini 2.0 Flash-Lite

text

Cost-optimized Gemini 2.0.

Google1.0M contextUS$0.075/M inputUS$0.3/M output

Google: Gemini 2.5 Flash

text

Best price-performance Gemini with thinking, for high-volume, low-latency work.

Google1.0M contextUS$0.3/M inputUS$2.5/M output

Google: Gemini 2.5 Flash-Lite

text

The fastest, lowest-cost Gemini 2.5 model.

Google1.0M contextUS$0.1/M inputUS$0.4/M output

Google: Gemini 2.5 Pro

text

Google's most capable thinking model for complex reasoning and coding over a 1M-token context.

Google1.0M contextUS$1.25/M inputUS$10/M output

Google: Gemma 3 12B

text

Mid-size Gemma 3 open model from Google with a 128K context and strong multilingual quality for its size.

Google131k contextUS$0.09/M inputUS$0.29/M output

Google: Gemma 3 27B

text

The largest Gemma 3 open model from Google, with a 128K context and quality close to much larger models.

Google131k contextUS$0.23/M inputUS$0.38/M output

Google: Gemma 3 4B

text

Small, fast open model from Google's Gemma 3 family with a 128K context. Good for lightweight chat, extraction and classification.

Google131k contextUS$0.04/M inputUS$0.08/M output

Google: Gemma 4 26B A4B

text

Mixture-of-experts Gemma 4 (25.2B total, 3.8B active per token): cost-efficient reasoning and function calling over a 256K context.

Google262k contextUS$0.13/M inputUS$0.4/M output

Google: Gemma 4 31B

text

Google's 31B dense Gemma 4 model with built-in reasoning, native function calling and text + image input over a 256K context.

Google262k contextUS$0.14/M inputUS$0.4/M output

Google: Gemma 4 E2B

text

Compact Gemma 4 (2.3B effective parameters) for low-latency workloads, with reasoning and function calling over a 128K context.

Google131k contextUS$0.04/M inputUS$0.08/M output

Meta: Llama 3 70B Instruct

text

Meta's Llama 3 70B instruction-tuned model for general chat, reasoning and writing.

Meta128k contextUS$2.65/M inputUS$3.5/M output

Meta: Llama 3 8B Instruct

text

Meta's compact Llama 3 instruction-tuned model. Fast and inexpensive for simple chat and text tasks.

Meta128k contextUS$0.3/M inputUS$0.6/M output

Meta: Llama 3.1 70B Instruct

text

Llama 3.1 70B with a 128K context, strong multilingual chat, reasoning and coding.

Meta131k contextUS$0.72/M inputUS$0.72/M output

Meta: Llama 3.1 8B Instruct

text

Llama 3.1 8B with a 128K context: a fast, low-cost open model for everyday tasks and high-volume workloads.

Meta131k contextUS$0.22/M inputUS$0.22/M output

Meta: Llama 3.3 70B Instruct

text

Multilingual 70B instruction-tuned model with performance close to much larger models on reasoning, math and general knowledge.

Meta131k contextUS$0.72/M inputUS$0.72/M output

Meta: Llama 4 Maverick 17B

text

Mixture-of-experts multimodal model with 128 experts and a 1M-token context window, tuned for assistant and chat use cases.

Meta1.0M contextUS$0.24/M inputUS$0.97/M output

Meta: Llama 4 Scout

text

Llama 4 Scout 17B Instruct (16E) is a mixture-of-experts (MoE) language model developed by Meta, activating 17 billion parameters out of a total of 109B.

Meta1.3M contextUS$0.17/M inputUS$0.66/M output

MiniMax: M2

text

MiniMax-M2 is a compact, high-efficiency large language model optimized for end-to-end coding and agentic workflows.

MiniMax205k contextUS$0.3/M inputUS$1.2/M output

MiniMax: M2.1

text

MiniMax-M2.1 is a lightweight, state-of-the-art large language model optimized for coding, agentic workflows, and modern application development.

MiniMax205k contextUS$0.3/M inputUS$1.2/M output

MiniMax: M2.5

text

MiniMax-M2.5 is a SOTA large language model designed for real-world productivity.

MiniMax205k contextUS$0.3/M inputUS$1.2/M output

Mistral: Codestral

text

Mistral's coding model for completion and fill-in-the-middle.

Mistral256k contextUS$0.3/M inputUS$0.9/M output

Mistral: Devstral 2 123B

text

Devstral 2 is a state-of-the-art open-source model by Mistral AI specializing in agentic coding.

Mistral262k contextUS$0.4/M inputUS$2/M output

Mistral: Devstral Medium

text

Agentic coding model for software engineering.

Mistral128k contextUS$0.4/M inputUS$2/M output

Mistral: Devstral Small

text

Small open agentic coding model.

Mistral128k contextUS$0.1/M inputUS$0.3/M output

Mistral: Magistral Medium

text

Mistral's reasoning model.

Mistral40k contextUS$2/M inputUS$5/M output

Mistral: Magistral Small 1.2

text

Mistral's open reasoning model (Apache 2.0) that thinks step by step before answering.

Mistral128k contextUS$0.5/M inputUS$1.5/M output

Mistral: Ministral 3 14B

text

The largest model in the Ministral 3 family, Ministral 3 14B offers frontier capabilities and performance comparable to its larger Mistral Small 3.2 24B counterpart.

Mistral262k contextUS$0.2/M inputUS$0.2/M output

Mistral: Ministral 3 3B

text

The smallest model in the Ministral 3 family, Ministral 3 3B is a powerful, efficient tiny language model with vision capabilities.

Mistral131k contextUS$0.1/M inputUS$0.1/M output

Mistral: Ministral 3 8B

text

A balanced model in the Ministral 3 family, Ministral 3 8B is a powerful, efficient tiny language model with vision capabilities.

Mistral262k contextUS$0.15/M inputUS$0.15/M output

Mistral: Ministral 3B

text

Smallest edge model.

Mistral128k contextUS$0.1/M inputUS$0.1/M output

Mistral: Ministral 8B

text

Edge-sized model for low-latency tasks.

Mistral128k contextUS$0.15/M inputUS$0.15/M output

Mistral: Mistral 7B Instruct

text

Mistral's compact 7B instruction model (Apache 2.0). Very fast and inexpensive for simple tasks.

Mistral128k contextUS$0.15/M inputUS$0.2/M output

Mistral: Mistral Large (24.02)

text

Mistral Large (February 2024): Mistral's earlier flagship for complex reasoning and multilingual work.

Mistral128k contextUS$4/M inputUS$12/M output

Mistral: Mistral Large 3

text

Mistral's open-weight flagship mixture-of-experts model (Apache 2.0) for reasoning, coding and long documents.

Mistral128k contextUS$0.5/M inputUS$1.5/M output

Mistral: Mistral Medium 3

text

Frontier-class multimodal model at a medium price.

Mistral128k contextUS$1.5/M inputUS$7.5/M output

Mistral: Mistral Small (24.02)

text

Mistral Small (February 2024): efficient model for low-latency, high-volume tasks.

Mistral128k contextUS$1/M inputUS$3/M output

Mistral: Mistral Small 3.2

text

Efficient small multimodal model.

Mistral128k contextUS$0.15/M inputUS$0.6/M output

Mistral: Mixtral 8x7B Instruct

text

Sparse mixture-of-experts model from Mistral (Apache 2.0) with good quality at low cost.

Mistral128k contextUS$0.45/M inputUS$0.7/M output

Mistral: Pixtral Large

text

124B multimodal model with frontier-level image understanding, strong document and chart analysis, and long-context reasoning.

Mistral128k contextUS$2/M inputUS$6/M output

Mistral: Voxtral Mini 3B

text

Compact Mistral model built for speech understanding; served here for text chat.

Mistral128k contextUS$0.04/M inputUS$0.04/M output

Mistral: Voxtral Small 24B

text

Mistral Small 3 variant with audio understanding; served here for text chat.

Mistral33k contextUS$0.1/M inputUS$0.3/M output

Moonshot AI: Kimi K2 Thinking

text

Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning.

Moonshot AI262k contextUS$0.6/M inputUS$2.5/M output

Moonshot AI: Kimi K2.5

text

Kimi K2.5 is Moonshot AI's native multimodal model, delivering state-of-the-art visual coding capability and a self-directed agent swarm paradigm.

Moonshot AI262k contextUS$0.6/M inputUS$3/M output

Moonshot AI: Kimi K3

text

Kimi K3 is a 2.8T parameter open-weight multimodal reasoning model from Moonshot AI.

Moonshot AI1.0M contextUS$3.3/M inputUS$16.5/M output

NVIDIA: Nemotron 3 Super 120B

text

NVIDIA Nemotron 3 Super is a 120B-parameter open hybrid MoE model, activating just 12B parameters for maximum compute efficiency and accuracy in complex multi-agent applications.

NVIDIA262k contextUS$0.15/M inputUS$0.65/M output

NVIDIA: Nemotron Nano 12B v2 VL

text

NVIDIA's compact vision-language Nemotron model; served here for text chat.

NVIDIA128k contextUS$0.2/M inputUS$0.6/M output

NVIDIA: Nemotron Nano 3 30B

text

NVIDIA Nemotron 3 Nano: efficient mixture-of-experts model for agents, chat and reasoning.

NVIDIA128k contextUS$0.06/M inputUS$0.24/M output

NVIDIA: Nemotron Nano 9B v2

text

NVIDIA's small hybrid reasoning model, tuned for fast, accurate answers with optional reasoning.

NVIDIA128k contextUS$0.06/M inputUS$0.23/M output

OpenAI: GPT-3.5 Turbo

text

Legacy low-cost chat model.

OpenAI16k contextUS$0.5/M inputUS$1.5/M output

OpenAI: GPT-4 Turbo

text

Previous-generation GPT-4 with vision.

OpenAI128k contextUS$10/M inputUS$30/M output

OpenAI: GPT-4.1

text

Non-reasoning model with strong instruction following and a 1M-token context.

OpenAI1.0M contextUS$2/M inputUS$8/M output

OpenAI: GPT-4.1 mini

text

Smaller, faster GPT-4.1 with the same 1M-token context.

OpenAI1.0M contextUS$0.4/M inputUS$1.6/M output

OpenAI: GPT-4.1 nano

text

The fastest, cheapest GPT-4.1.

OpenAI1.0M contextUS$0.1/M inputUS$0.4/M output

OpenAI: GPT-4o

text

Versatile multimodal model for text and image input.

OpenAI128k contextUS$2.5/M inputUS$10/M output

OpenAI: GPT-4o mini

text

Low-cost multimodal model for high-volume tasks.

OpenAI128k contextUS$0.15/M inputUS$0.6/M output

OpenAI: GPT-5

text

OpenAI's flagship reasoning model for coding, agents and complex multi-step work.

OpenAI400k contextUS$1.25/M inputUS$10/M output

OpenAI: GPT-5 Chat

text

The non-reasoning GPT-5 used in ChatGPT.

OpenAI128k contextUS$1.25/M inputUS$10/M output

OpenAI: GPT-5 Codex

text

GPT-5 tuned for agentic coding.

OpenAI400k contextUS$1.25/M inputUS$10/M output

OpenAI: GPT-5 mini

text

A faster, cheaper GPT-5 for well-defined tasks.

OpenAI400k contextUS$0.25/M inputUS$2/M output

OpenAI: GPT-5 nano

text

The fastest, cheapest GPT-5, for classification and summarization.

OpenAI400k contextUS$0.05/M inputUS$0.4/M output

OpenAI: GPT-6 Luna

text

OpenAI's most efficient model for focused, high-volume tasks.

OpenAI1.1M contextUS$0.1/M inputUS$0.5/M output

OpenAI: GPT-6.1 Terra

text

A fast, efficient model for visual reasoning, precise computer use, and clear, natural writing.

OpenAI1.1M contextUS$1/M inputUS$5/M output

OpenAI: GPT-6.2 Astra

text

OpenAI's most capable model for the hardest projects: software, computer use and documents.

OpenAI1.1M contextUS$8/M inputUS$40/M output

OpenAI: GPT-6.2 Sol

text

OpenAI's go-to model for everyday work, from coding to powering agents at scale.

OpenAI1.1M contextUS$2/M inputUS$10/M output

OpenAI: gpt-oss-120b

text

OpenAI's open-weight 117B mixture-of-experts reasoning model. Available through Bring Your Own Key with your AWS account.

OpenAI131k contextUS$0.15/M inputUS$0.6/M output

OpenAI: gpt-oss-20b

text

gpt-oss-20b is an open-weight 21B parameter model released by OpenAI under the Apache 2.0 license.

OpenAI131k contextUS$0.07/M inputUS$0.3/M output

OpenAI: gpt-oss-safeguard-120b

text

OpenAI safety reasoning model built on gpt-oss-120b that classifies content against a policy you provide.

OpenAI128k contextUS$0.15/M inputUS$0.6/M output

OpenAI: gpt-oss-safeguard-20b

text

gpt-oss-safeguard-20b is a safety reasoning model from OpenAI built upon gpt-oss-20b.

OpenAI131k contextUS$0.07/M inputUS$0.2/M output

OpenAI: o1

text

Previous full reasoning model.

OpenAI200k contextUS$15/M inputUS$60/M output

OpenAI: o1-pro

text

o1 with more compute (Responses API).

OpenAI200k contextUS$150/M inputUS$600/M output

OpenAI: o3

text

Reasoning model for math, science, coding and visual reasoning.

OpenAI200k contextUS$2/M inputUS$8/M output

OpenAI: o3-mini

text

Small reasoning model for STEM tasks.

OpenAI200k contextUS$1.1/M inputUS$4.4/M output

OpenAI: o3-pro

text

o3 with more compute for harder problems.

OpenAI200k contextUS$20/M inputUS$80/M output

OpenAI: o4-mini

text

Fast, cost-efficient reasoning model.

OpenAI200k contextUS$1.1/M inputUS$4.4/M output

Qwen: Qwen3 32B

text

Dense 32B model with hybrid thinking, solid multilingual ability and good instruction following at a low price.

Qwen131k contextUS$0.15/M inputUS$0.6/M output

Qwen: Qwen3 Coder 30B A3B

text

Efficient mixture-of-experts coding model with 3B active parameters, built for repository-scale code understanding and agentic coding.

Qwen262k contextUS$0.15/M inputUS$0.6/M output

Qwen: Qwen3 Coder Next

text

Qwen3-Coder-Next is an open-weight causal language model optimized for coding agents and local development workflows.

Qwen262k contextUS$0.5/M inputUS$1.2/M output

Qwen: Qwen3 Next 80B A3B

text

Qwen3-Next-80B-A3B-Instruct is an instruction-tuned chat model in the Qwen3-Next series optimized for fast, stable responses without “thinking” traces.

Qwen262k contextUS$0.14/M inputUS$1.2/M output

Qwen: Qwen3 VL 235B A22B

text

Qwen3-VL-235B-A22B Instruct is an open-weight multimodal model that unifies strong text generation with visual understanding across images and video.

Qwen262k contextUS$0.53/M inputUS$2.66/M output

Writer: Palmyra X4

text

Writer's enterprise model for business writing, agents and tool use.

Writer128k contextUS$2.5/M inputUS$10/M output

Writer: Palmyra X5

text

Palmyra X5 is Writer's most advanced model, purpose-built for building and scaling AI agents across the enterprise.

Writer1.0M contextUS$0.6/M inputUS$6/M output

xAI: Grok 4.6

text

xAI's Grok 4.6 with frontier performance on coding, knowledge work and STEM.

xAI500k contextUS$2.2/M inputUS$6.6/M output

Z.AI: GLM 4.7

text

GLM-4.7 is Z.ai’s latest flagship model, featuring upgrades in two key areas: enhanced programming capabilities and more stable multi-step reasoning/execution.

Z.AI205k contextUS$0.6/M inputUS$2.2/M output

Z.AI: GLM 4.7 Flash

text

As a 30B-class SOTA model, GLM-4.7-Flash offers a new option that balances performance and efficiency.

Z.AI200k contextUS$0.07/M inputUS$0.4/M output

Z.AI: GLM 5

text

Z.AI's GLM 5 flagship for coding, agents and long multi-step reasoning.

Z.AI128k contextUS$1/M inputUS$3.2/M output