Provider Snapshot
13
64.12
722.31
May 7, 2026
Key Takeaways
13 bedrock models are actively benchmarked with 17137 total measurements across 14807 benchmark runs.
nova-micro leads the fleet with 125.00 tokens/second, while nova-pro delivers 85.60 tok/s.
Performance varies by 46.0% across the bedrock model lineup, indicating diverse optimization strategies for different use cases.
Avg time to first token across the fleet is 722.31 ms, showing good responsiveness for interactive applications.
The bedrock model fleet shows varied performance characteristics (55.0% variation coefficient), reflecting diverse model architectures.
Fastest Models
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| bedrock | nova-micro | 125.00 | 8.55 | 151.00 | 270.00 |
| bedrock | llama-4-maverick | 107.00 | 1.03 | 139.00 | 440.00 |
| bedrock | nova-lite | 98.80 | 8.33 | 132.00 | 310.00 |
| bedrock | llama-4-scout | 94.20 | 1.09 | 130.00 | 340.00 |
| bedrock | llama-3.3-70b | 90.70 | 3.81 | 122.00 | 320.00 |
| bedrock | nova-pro | 85.60 | 8.15 | 123.00 | 370.00 |
All Models
Complete list of all bedrock models tracked in the benchmark system. Click any model name to view detailed performance data.
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| bedrock | nova-lite | 98.80 | 8.33 | 132.00 | 310.00 |
| bedrock | nova-micro | 125.00 | 8.55 | 151.00 | 270.00 |
| bedrock | nova-pro | 85.60 | 8.15 | 123.00 | 370.00 |
| bedrock | mistral-large | 40.70 | 2.50 | 47.00 | 610.00 |
| bedrock | claude-3-5-haiku | 32.50 | 2.66 | 41.00 | 700.00 |
| bedrock | claude-3-7-sonnet | 31.20 | 1.07 | 42.30 | 1100.00 |
| bedrock | claude-haiku-4.5 | 41.70 | 2.11 | 61.90 | 1070.00 |
| bedrock | claude-opus-4.5 | 19.50 | 2.12 | 23.90 | 1850.00 |
| bedrock | claude-sonnet-4.5 | 21.10 | 2.53 | 28.10 | 1590.00 |
| bedrock | llama-3.2-90b | 45.50 | 3.59 | 50.10 | 420.00 |
| bedrock | llama-3.3-70b | 90.70 | 3.81 | 122.00 | 320.00 |
| bedrock | llama-4-maverick | 107.00 | 1.03 | 139.00 | 440.00 |
| bedrock | llama-4-scout | 94.20 | 1.09 | 130.00 | 340.00 |
Featured Models
Frequently Asked Questions
Based on recent tests, nova-micro shows the highest average throughput among tracked bedrock models.
This provider summary aggregates 17137 individual prompts measured across 14807 monitoring runs over the past month.