Provider Snapshot
9
24.91
1.18
Dec 26, 2025
Key Takeaways
9 anthropic models are actively benchmarked with 7756 total measurements across 7756 benchmark runs.
claude-haiku-4.5 leads the fleet with 50.50 tokens/second, while claude-opus-4.5 delivers 20.20 tok/s.
Performance varies by 150.0% across the anthropic model lineup, indicating diverse optimization strategies for different use cases.
Median time to first token across the fleet is 1.18 ms, showing excellent responsiveness for interactive applications.
The anthropic model fleet shows varied performance characteristics (41.3% variation coefficient), reflecting diverse model architectures.
Fastest Models
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| anthropic | claude-haiku-4.5 | 50.50 | 8.01 | 71.80 | 0.56 |
| anthropic | claude-3-7-sonnet | 30.60 | 8.91 | 43.50 | 0.59 |
| anthropic | claude-3-5-haiku | 30.20 | 5.82 | 40.40 | 0.54 |
| anthropic | claude-4-sonnet | 20.50 | 4.11 | 32.40 | 1.50 |
| anthropic | claude-3-opus | 20.20 | 2.34 | 27.10 | 1.06 |
| anthropic | claude-opus-4.5 | 20.20 | 4.48 | 29.80 | 1.70 |
All Models
Complete list of all anthropic 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) |
|---|---|---|---|---|---|
| anthropic | claude-3-5-haiku | 30.20 | 5.82 | 40.40 | 0.54 |
| anthropic | claude-3-7-sonnet | 30.60 | 8.91 | 43.50 | 0.59 |
| anthropic | claude-3-opus | 20.20 | 2.34 | 27.10 | 1.06 |
| anthropic | claude-4-opus | 17.30 | 7.17 | 23.80 | 1.24 |
| anthropic | claude-4-sonnet | 20.50 | 4.11 | 32.40 | 1.50 |
| anthropic | Claude Opus 4.1 | 17.40 | 6.32 | 24.10 | 1.43 |
| anthropic | claude-sonnet-4.5 | 17.30 | 4.21 | 23.30 | 1.97 |
| anthropic | claude-opus-4.5 | 20.20 | 4.48 | 29.80 | 1.70 |
| anthropic | claude-haiku-4.5 | 50.50 | 8.01 | 71.80 | 0.56 |
Featured Models
Frequently Asked Questions
Based on recent tests, claude-haiku-4.5 shows the highest average throughput among tracked anthropic models.
This provider summary aggregates 7756 individual prompts measured across 7756 monitoring runs over the past month.