Amazon Bedrock: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by Amazon Bedrock for your use-case.
Most Intelligent
Claude Opus 5.5 (max with fallback)Claude Sonnet 5.5 (max with fallback)
GPT-6 Astra (max)Claude Opus 5 (max)GPT-5.6 Sol (max)Intelligence Index
Total 77 models
Fastest
Ministral 3 3BNova Micro
Grok 4.3 (high)
Qwen3 Coder 30B A3BGrok 4.3 (non-reasoning)Output speed
Total 77 models
Lowest Price
Nova Micro
Gemma 3 4BNova Lite
NVIDIA Nemotron Nano 9B V2 (non-reasoning)gpt-oss-20b (low)Blended price (per 1M tokens)
Total 77 models
Amazon offers 77 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For intelligence, the top models on Amazon are Claude Opus 5.5 (max with fallback) (58), Claude Sonnet 5.5 (max with fallback) (56), and GPT-6 Astra (max) (53).
- For output speed, the fastest models are Ministral 3 3B (422 t/s), Nova Micro (304 t/s), and Grok 4.3 (high) (280 t/s). Speed varies significantly across models, with a 66% difference between the fastest and slowest.
- For latency, GPT-5.6 Luna (non-reasoning) (0.43s), GPT-5.6 Terra (non-reasoning) (0.51s), and GPT-5.6 Sol (non-reasoning) (0.61s) offer the lowest time to first answer token.
- For pricing, Nova Micro ($0.03), Gemma 3 4B ($0.04), and Nova Lite ($0.05) offer the lowest blended prices per 1M tokens. Prices vary up to 3.4x across models.
- For context window size, Claude Opus 5.5 (max with fallback) (1M), Claude Sonnet 5.5 (max with fallback) (1M), and Claude Opus 5 (max) (1M) support the largest context windows on Amazon.
Intelligence Evaluations
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic real-world work tasks, Elo
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Long context reasoning
Coding
Physics reasoning
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic interactive 3D experiences, Elo
Product and brand videos designed in code, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Context Window
Context Window
Pricing
Intelligence Index vs. Price
Performance Summary
Output Speed vs. Price
Speed
Measured by Output Speed (tokens per second)
Output Speed
Latency
Measured by Time (seconds) to First Token
Latency: Time To First Answer Token
Cache Behaviour
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
End-to-End Response Time
Seconds to output 500 tokens, calculated based on time to first token, 'thinking' time for reasoning models, and output speed
End-to-End Response Time vs. Price
Further Analysis | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
Claude Opus 5.5 (max with fallback) | 1M | Proprietary | 58 | $5.98 | 98 | 748.89 | 753.99 | -- | |||
Claude Sonnet 5.5 (max with fallback) | 1M | Proprietary | 56 | $4.96 | 143 | 436.35 | 439.84 | -- | |||
GPT-6 Astra (max) | 524k | Proprietary | 53 | $3.26 | 91 | 314.97 | 320.47 | -- | |||
Claude Opus 5 (max) | 1M | Proprietary | 51 | $5.86 | 57 | 59.83 | 68.58 | -- | |||
Claude Opus 5 (xhigh) | 1M | Proprietary | 50 | $4.88 | -- | -- | -- | -- | |||
Claude Opus 5 (high) | 1M | Proprietary | 48 | $3.61 | -- | -- | -- | -- | |||
GPT-5.6 Sol (max) | 1M | Proprietary | 47 | $1.98 | 130 | 69.51 | 73.37 | -- | |||
Claude Opus 5 (medium) | 1M | Proprietary | 45 | $2.19 | -- | -- | -- | -- | |||
 Kimi K3 (max) | 1M | Open | 44 | $1.87 | 220 | 1.23 | 12.58 | 9.07 | |||
Claude Opus 5.5 (low with fallback) | 1M | Proprietary | 42 | $0.55 | 81 | 11.04 | 17.19 | -- | |||
GPT-5.6 Terra (max) | 1M | Proprietary | 42 | $1.41 | 158 | 116.38 | 119.55 | -- | |||
Claude Opus 4.8 (max) | 1M | Proprietary | 42 | $4.02 | 68 | 35.69 | 43.07 | -- | |||
Claude Opus 4.7 (max) | 1M | Proprietary | 41* | -- | 70 | 17.21 | 24.34 | -- | |||
Claude Opus 5 (low) | 1M | Proprietary | 39 | $1.10 | 55 | 2.33 | 11.46 | -- | |||
GPT-5.4 (xhigh) | 272k | Proprietary | 39* | -- | 189 | 93.83 | 96.48 | -- | |||
GPT-5.5 (xhigh) | 272k | Proprietary | 38 | $2.49 | 138 | 62.19 | 65.82 | -- | |||
Claude Sonnet 5 (max) | 1M | Proprietary | 38 | $6.21 | 90 | 200.92 | 206.46 | -- | |||
GPT-5.6 Luna (max) | 1M | Proprietary | 37 | $0.18 | 224 | 77.62 | 79.85 | -- | |||
GPT-5.5 (high) | 272k | Proprietary | 37 | $1.40 | -- | -- | -- | -- | |||
Claude Sonnet 5.5 (low with fallback) | 1M | Proprietary | 36 | $0.33 | 107 | 1.56 | 6.22 | -- | |||
GPT-5.5 (medium) | 272k | Proprietary | 34 | $0.92 | -- | -- | -- | -- | |||
Claude Opus 4.6 (max) | 1M | Proprietary | 32* | -- | 68 | 17.48 | 24.86 | -- | |||
Claude Opus 4.7 (non-reasoning, high) | 1M | Proprietary | 31* | -- | 66 | 1.13 | 8.71 | -- | |||
GPT-5.5 (low) | 272k | Proprietary | 31* | -- | -- | -- | -- | -- | |||
Claude Sonnet 4.6 (max) | 200k | Proprietary | 30 | $2.83 | 116 | 58.14 | 62.45 | -- | |||
Claude Opus 4.5 | 200k | Proprietary | 29* | -- | 60 | 16.94 | 25.29 | -- | |||
GPT-5.6 Sol (non-reasoning) | 1M | Proprietary | 28* | -- | 113 | 0.61 | 5.02 | -- | |||
GPT-5.4 (low) | 272k | Proprietary | 28* | -- | -- | -- | -- | -- | |||
Claude Opus 4.6 (non-reasoning, high) | 1M | Proprietary | 26* | -- | 62 | 1.31 | 9.34 | -- | |||
Grok 4.3 (high) | 524k | Proprietary | 25 | $0.34 | 280 | 9.81 | 11.59 | -- | |||
Grok 4.3 (medium) | 524k | Proprietary | 25* | -- | -- | -- | -- | -- | |||
Claude Sonnet 4.6 (non-reasoning, high) | 200k | Proprietary | 25* | -- | 96 | 1.09 | 6.29 | -- | |||
Grok 4.3 (low) | 524k | Proprietary | 24* | -- | -- | -- | -- | -- | |||
Claude Opus 4.5 (non-reasoning) | 200k | Proprietary | 24* | -- | 56 | 1.48 | 10.40 | -- | |||
Kimi K2.5 | 256k | Open | 23* | -- | 99 | 1.24 | 36.21 | 29.93 | |||
Claude Sonnet 5 (non-reasoning) | 1M | Proprietary | 23* | -- | 62 | 1.24 | 9.33 | -- | |||
GPT-5.5 (non-reasoning) | 272k | Proprietary | 23* | -- | 115 | 0.80 | 5.13 | -- | |||
Claude 4.1 Opus | 200k | Proprietary | 23* | -- | -- | -- | -- | -- | |||
 GLM-4.7 | 200k | Open | 22* | -- | 75 | 1.41 | 34.59 | 26.54 | |||
Kimi K2 Thinking | 256k | Open | 22* | -- | 121 | 1.35 | 21.96 | 16.49 | |||
 DeepSeek V3.2 | 128k | Open | 21* | -- | 92 | 1.55 | 28.72 | 21.74 | |||
GPT-5.6 Terra (non-reasoning) | 1M | Proprietary | 21 | $0.14 | 126 | 0.51 | 4.48 | -- | |||
Claude 4.5 Sonnet | 1M | Proprietary | 21 | $0.55 | 60 | 12.59 | 20.98 | -- | |||
Claude 4.5 Sonnet (non-reasoning) | 1M | Proprietary | 19* | -- | 55 | 1.41 | 10.55 | -- | |||
 MiniMax-M2 | 205k | Open | 19* | -- | 128 | 1.04 | 20.63 | 15.68 | |||
Claude 4.1 Opus (non-reasoning) | 200k | Proprietary | 19* | -- | -- | -- | -- | -- | |||
GPT-5.4 (non-reasoning) | 272k | Proprietary | 18* | -- | 141 | 0.73 | 4.27 | -- | |||
GLM-4.7 (non-reasoning) | 200k | Open | 17* | -- | 75 | 1.40 | 8.10 | -- | |||
Claude 4.5 Haiku | 200k | Proprietary | 17 | $0.22 | 139 | 21.35 | 24.94 | -- | |||
DeepSeek V3.2 (non-reasoning) | 128k | Open | 16* | -- | 93 | 1.49 | 6.88 | -- | |||
GPT-5.6 Luna (non-reasoning) | 1M | Proprietary | 16 | $0.01 | 213 | 0.43 | 2.78 | -- | |||
Claude 4.5 Haiku (non-reasoning) | 200k | Proprietary | 15* | -- | 109 | 0.84 | 5.44 | -- | |||
GLM-4.7-Flash | 200k | Open | 15* | -- | 221 | 1.03 | 12.36 | 9.06 | |||
Nova 2.0 Pro Preview (medium) | 256k | Proprietary | 14* | -- | 129 | 14.15 | 33.60 | 15.55 | |||
Grok 4.3 (non-reasoning) | 524k | Proprietary | 14 | $0.22 | 263 | 0.80 | 2.70 | -- | |||
DeepSeek V3.1 (non-reasoning) | 128k | Open | 14* | -- | 159 | 1.31 | 4.45 | -- | |||
DeepSeek V3.1 | 128k | Open | 13* | -- | 166 | 1.30 | 16.34 | 12.03 | |||
Nova 2.0 Lite (high) | 1M | Proprietary | 13* | -- | 209 | 17.14 | 29.12 | 9.59 | |||
Nova 2.0 Pro Preview (low) | 256k | Proprietary | 13* | -- | 124 | 10.03 | 30.25 | 16.17 | |||
Nova 2.0 Lite (medium) | 1M | Proprietary | 12* | -- | 195 | 16.93 | 29.73 | 10.24 | |||
Qwen3 235B 2507 | 256k | Open | 12* | -- | 107 | 0.98 | 5.64 | -- | |||
Qwen3 Coder 480B | 262k | Open | 12* | -- | 87 | 1.17 | 6.91 | -- | |||
Nova 2.0 Lite (low) | 1M | Proprietary | 12* | -- | 183 | 10.27 | 23.91 | 10.92 | |||
gpt-oss-120b (high) | 131k | Open | 12 | $0.11 | 109 | 0.96 | 23.97 | 18.41 | |||
DeepSeek R1 (Jan) | 128k | Open | 11 | $0.22 | 161 | 1.03 | 18.69 | 14.56 | |||
GLM-4.7-Flash (non-reasoning) | 200k | Open | 11* | -- | 225 | 1.01 | 3.23 | -- | |||
gpt-oss-120b (low) | 131k | Open | 10* | -- | 96 | 1.00 | 26.99 | 20.79 | |||
Nova 2.0 Pro Preview (non-reasoning) | 256k | Proprietary | 10* | -- | 121 | 1.07 | 5.21 | -- | |||
gpt-oss-20b (low) | 131k | Open | 10* | -- | 200 | 6.50 | 19.03 | 10.02 | |||
Qwen3 Coder 30B A3B | 262k | Open | 10* | -- | 268 | 0.91 | 2.77 | -- | |||
 Llama 4 Maverick | 128k | Open | 9 | $0.04 | -- | -- | -- | -- | |||
Mistral Large 3 | 256k | Open | 9 | $0.10 | 146 | 1.23 | 4.65 | -- | |||
Qwen3 Coder Next | 128k | Open | 9 | $0.78 | 125 | 0.95 | 4.94 | -- | |||
gpt-oss-20b (high) | 131k | Open | 9 | $0.01 | 243 | 26.11 | 36.41 | 8.24 | |||
Nova 2.0 Lite (non-reasoning) | 1M | Proprietary | 9* | -- | 180 | 0.96 | 3.73 | -- | |||
Magistral Small 1.2 | 128k | Open | 9* | -- | 77 | 1.45 | 33.82 | 25.90 | |||
Claude 3.5 Sonnet (Oct) | 200k | Proprietary | 8* | -- | -- | -- | -- | -- | |||
Llama 3.3 70B | 128k | Open | 8* | -- | -- | -- | -- | -- | |||
Qwen3 32B (non-reasoning) | 32.8k | Open | 7* | -- | 144 | 1.06 | 4.54 | -- | |||
Claude 3.5 Sonnet (June) | 200k | Proprietary | 7* | -- | -- | -- | -- | -- | |||
Nova Pro | 300k | Proprietary | 7* | -- | 126 | 1.04 | 4.99 | -- | |||
Llama 3.1 8B | 128k | Open | 7* | -- | -- | -- | -- | -- | |||
NVIDIA Nemotron Nano 9B V2 (non-reasoning) | 131k | Open | 7* | -- | 170 | 1.06 | 4.00 | -- | |||
Mistral Large 2 (Jul) | 128k | Open | 7* | -- | -- | -- | -- | -- | |||
Nova Lite | 300k | Proprietary | 7* | -- | 174 | 0.88 | 3.75 | -- | |||
Llama 3.1 70B Standard | 128k | Open | 7* | -- | -- | -- | -- | -- | |||
Llama 3.1 70B Latency Optimized | 128k | Open | 7* | -- | -- | -- | -- | -- | |||
Llama 4 Scout | 128k | Open | 6 | $0.11 | -- | -- | -- | -- | |||
Ministral 3 14B | 256k | Open | 6 | $0.14 | 190 | 1.03 | 3.67 | -- | |||
 Jamba 1.5 Large | 256k | Open | 6* | -- | -- | -- | -- | -- | |||
Nova Micro | 130k | Proprietary | 6* | -- | 305 | 0.80 | 2.44 | -- | |||
NVIDIA Nemotron Nano 12B v2 VL (non-reasoning) | 128k | Open | 6* | -- | 202 | 1.03 | 3.50 | -- | |||
Mistral Large (Feb) | 32.8k | Proprietary | 6* | -- | -- | -- | -- | -- | |||
Claude 3 Haiku | 200k | Proprietary | 6* | -- | -- | -- | -- | -- | |||
Ministral 3 8B | 256k | Open | 5 | $0.07 | 251 | 0.94 | 2.93 | -- | |||
Llama 3 70B | 8.19k | Open | 5* | -- | -- | -- | -- | -- | |||
Jamba 1.5 Mini | 256k | Open | 5* | -- | -- | -- | -- | -- | |||
Mixtral 8x7B | 32.8k | Open | 5* | -- | 25 | 1.04 | 21.02 | -- | |||
Mistral 7B | 8.19k | Open | 5* | -- | -- | -- | -- | -- | |||
Gemma 3 27B | 128k | Open | 5 | $0.47 | 58 | 1.40 | 9.98 | -- | |||
Ministral 3 3B | 256k | Open | 5 | $0.04 | 422 | 0.85 | 2.04 | -- | |||
Gemma 3 4B | 128k | Open | 5* | -- | 174 | 0.97 | 3.84 | -- | |||
Llama 3 8B | 8.19k | Open | 5* | -- | -- | -- | -- | -- | |||
Gemma 3 12B | 128k | Open | 4 | $0.24 | 88 | 1.32 | 7.00 | -- | |||
Key definitions
Frequently Asked Questions
Common questions about Amazon Bedrock
Amazon Bedrock offers 77 models that we track: Claude Opus 5.5 (max with fallback), Claude Sonnet 5.5 (max with fallback), GPT-6 Astra (max), Claude Opus 5 (max), GPT-5.6 Sol (max), Kimi K3 (max), Claude Opus 5.5 (low with fallback), GPT-5.6 Terra (max), Claude Opus 4.8 (max), Claude Opus 4.7 (max), Claude Opus 5 (low), GPT-5.4 (xhigh), GPT-5.5 (xhigh), Claude Sonnet 5 (max), GPT-5.6 Luna (max), Claude Sonnet 5.5 (low with fallback), Claude Opus 4.6 (max), Claude Opus 4.7 (non-reasoning, high), Claude Sonnet 4.6 (max), Claude Opus 4.5, GPT-5.6 Sol (non-reasoning), Claude Opus 4.6 (non-reasoning, high), Grok 4.3 (high), Claude Sonnet 4.6 (non-reasoning, high), Claude Opus 4.5 (non-reasoning), Kimi K2.5, Claude Sonnet 5 (non-reasoning), GPT-5.5 (non-reasoning), GLM-4.7, Kimi K2 Thinking, DeepSeek V3.2, GPT-5.6 Terra (non-reasoning), Claude 4.5 Sonnet, Claude 4.5 Sonnet (non-reasoning), MiniMax-M2, GPT-5.4 (non-reasoning), GLM-4.7 (non-reasoning), Claude 4.5 Haiku, DeepSeek V3.2 (non-reasoning), GPT-5.6 Luna (non-reasoning), Claude 4.5 Haiku (non-reasoning), GLM-4.7-Flash, Nova 2.0 Pro Preview (medium), Grok 4.3 (non-reasoning), DeepSeek V3.1 (non-reasoning), DeepSeek V3.1, Nova 2.0 Lite (high), Nova 2.0 Pro Preview (low), Nova 2.0 Lite (medium), Qwen3 235B 2507, Qwen3 Coder 480B, Nova 2.0 Lite (low), gpt-oss-120b (high), DeepSeek R1 (Jan), GLM-4.7-Flash (non-reasoning), gpt-oss-120b (low), Nova 2.0 Pro Preview (non-reasoning), gpt-oss-20b (low), Qwen3 Coder 30B A3B, Mistral Large 3, Qwen3 Coder Next, gpt-oss-20b (high), Nova 2.0 Lite (non-reasoning), Magistral Small 1.2, Qwen3 32B (non-reasoning), Nova Pro, NVIDIA Nemotron Nano 9B V2 (non-reasoning), Nova Lite, Ministral 3 14B, Nova Micro, NVIDIA Nemotron Nano 12B v2 VL (non-reasoning), Ministral 3 8B, Mixtral 8x7B, Gemma 3 27B, Ministral 3 3B, Gemma 3 4B, and Gemma 3 12B.
The most intelligent model available on Amazon Bedrock is Claude Opus 5.5 (max with fallback) with an Intelligence Index score of 58.
The fastest model on Amazon Bedrock by output speed is Ministral 3 3B at 421.5 tokens per second.
The model with the lowest time to first answer token on Amazon Bedrock is GPT-5.6 Luna (non-reasoning) at 0.43s. Lower latency means faster initial response time.
The most affordable model on Amazon Bedrock by blended price is Nova Micro at $0.03 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Amazon Bedrock vary up to 284x across models, from $0.03 per 1M tokens for Nova Micro to $7.70 per 1M tokens for GPT-6 Astra (max).
18 of 77 models on Amazon Bedrock support JSON mode for structured output.
75 of 77 models on Amazon Bedrock support function calling (tool use).
Yes, Amazon Bedrock offers 40 reasoning models: Claude Opus 5.5 (max with fallback), Claude Sonnet 5.5 (max with fallback), GPT-6 Astra (max), Claude Opus 5 (max), GPT-5.6 Sol (max), Kimi K3 (max), Claude Opus 5.5 (low with fallback), GPT-5.6 Terra (max), Claude Opus 4.8 (max), Claude Opus 4.7 (max), Claude Opus 5 (low), GPT-5.4 (xhigh), GPT-5.5 (xhigh), Claude Sonnet 5 (max), GPT-5.6 Luna (max), Claude Sonnet 5.5 (low with fallback), Claude Opus 4.6 (max), Claude Sonnet 4.6 (max), Claude Opus 4.5, Grok 4.3 (high), Kimi K2.5, GLM-4.7, Kimi K2 Thinking, DeepSeek V3.2, Claude 4.5 Sonnet, MiniMax-M2, Claude 4.5 Haiku, GLM-4.7-Flash, Nova 2.0 Pro Preview (medium), DeepSeek V3.1, Nova 2.0 Lite (high), Nova 2.0 Pro Preview (low), Nova 2.0 Lite (medium), Nova 2.0 Lite (low), gpt-oss-120b (high), DeepSeek R1 (Jan), gpt-oss-120b (low), gpt-oss-20b (low), gpt-oss-20b (high), and Magistral Small 1.2. Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, 33 of 77 models on Amazon Bedrock are open weight models: Kimi K3 (max), Kimi K2.5, GLM-4.7, Kimi K2 Thinking, DeepSeek V3.2, MiniMax-M2, GLM-4.7 (non-reasoning), DeepSeek V3.2 (non-reasoning), GLM-4.7-Flash, DeepSeek V3.1 (non-reasoning), DeepSeek V3.1, Qwen3 235B 2507, Qwen3 Coder 480B, gpt-oss-120b (high), DeepSeek R1 (Jan), GLM-4.7-Flash (non-reasoning), gpt-oss-120b (low), gpt-oss-20b (low), Qwen3 Coder 30B A3B, Mistral Large 3, Qwen3 Coder Next, gpt-oss-20b (high), Magistral Small 1.2, Qwen3 32B (non-reasoning), NVIDIA Nemotron Nano 9B V2 (non-reasoning), Ministral 3 14B, NVIDIA Nemotron Nano 12B v2 VL (non-reasoning), Ministral 3 8B, Mixtral 8x7B, Gemma 3 27B, Ministral 3 3B, Gemma 3 4B, and Gemma 3 12B.
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on Amazon Bedrock, consider: intelligence (for quality-sensitive tasks), output speed (for throughput-intensive tasks), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.