llm-benchmarks
githubdrose.io

bedrock Provider Benchmarks


Measured over the last 30 days

Models Tracked
26
Avg Tokens / Second
59.34
Avg Time to First Token (ms)
735.77
Last Updated
Sep 23, 2026

All bedrock models

26 live · one row per model · fastest first · 2 retired lanes not listed · 7 folded behind their current lane
ModelAvg Toks/SecMinMaxAvg TTF (ms)
Nova Micro 1.0117.0011.20169.00320.00
Nova Lite 1.0106.0013.70146.00330.00
Llama 4 Maverick106.0023.70131.00290.00
Llama 3.1 8B Instruct95.808.97110.00400.00
Llama 4 Scout95.601.05120.00340.00
Llama 3.3 70B Instruct89.405.38127.00300.00
Nova Pro 1.086.901.76131.00410.00
mistral-7b82.1025.0089.80200.00
llama-3-8b77.8010.9086.00230.00
mixtral-8x7b75.0021.5080.90220.00
Kimi K2.571.0069.0072.90500.00
Llama 3.1 70B Instruct60.301.9573.80440.00
mistral-small56.7015.1060.00210.00
Nova 2 Lite53.3043.2058.70710.00
Claude Haiku 4.548.702.8864.60790.00
Palmyra X543.8034.9050.20720.00
Mistral Large42.805.1146.80260.00
llama-3-70b38.804.9541.70300.00
Claude Haiku Latest38.4037.4039.90900.00
Claude Opus 4.728.604.6437.501390.00
Claude Sonnet 4.628.401.0334.801000.00
Nova Premier 1.027.1026.3027.70670.00
Claude Opus 4.622.006.4226.801570.00
Claude Sonnet 4.521.902.2826.101410.00
Claude Opus 4.520.901.2225.301640.00
Claude Opus 4.18.461.1315.803580.00

Frequently Asked Questions

Which bedrock model is fastest?
Based on recent tests, Nova Micro 1.0 shows the highest average throughput among tracked bedrock models.
How many recent measurements feed this dashboard?
This provider summary aggregates 29713 individual prompts measured across 26600 monitoring runs over the past month.