Provider Snapshot
8
25.95
1323.75
May 20, 2026
Key Takeaways
8 anthropic models are actively benchmarked with 1354 total measurements across 1299 benchmark runs.
claude-haiku-4.5 leads the fleet with 50.10 tokens/second, while claude-opus-4.6 delivers 21.20 tok/s.
Performance varies by 136.3% across the anthropic model lineup, indicating diverse optimization strategies for different use cases.
Avg time to first token across the fleet is 1323.75 ms, showing moderate responsiveness for interactive applications.
The anthropic model fleet shows varied performance characteristics (37.5% variation coefficient), reflecting diverse model architectures.
Fastest Models
| Provider | Model | Avg Toks/Sec | Min | Max | Avg TTF (ms) |
|---|---|---|---|---|---|
| anthropic | claude-haiku-4.5 | 50.10 | 10.60 | 65.40 | 720.00 |
| anthropic | claude-4-sonnet | 26.70 | 8.20 | 37.20 | 1610.00 |
| anthropic | claude-sonnet-4.6 | 25.40 | 11.80 | 34.60 | 1030.00 |
| anthropic | claude-opus-4.5 | 25.10 | 10.80 | 31.40 | 1140.00 |
| anthropic | claude-opus-4-7 | 24.20 | 7.88 | 31.60 | 1750.00 |
| anthropic | claude-opus-4.6 | 21.20 | 9.49 | 29.30 | 1500.00 |
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-haiku-4.5 | 50.10 | 10.60 | 65.40 | 720.00 |
| anthropic | Claude Opus 4.1 | 18.60 | 15.10 | 22.50 | 1330.00 |
| anthropic | claude-4-opus | 16.30 | 6.49 | 21.10 | 1510.00 |
| anthropic | claude-opus-4.5 | 25.10 | 10.80 | 31.40 | 1140.00 |
| anthropic | claude-opus-4.6 | 21.20 | 9.49 | 29.30 | 1500.00 |
| anthropic | claude-opus-4-7 | 24.20 | 7.88 | 31.60 | 1750.00 |
| anthropic | claude-4-sonnet | 26.70 | 8.20 | 37.20 | 1610.00 |
| anthropic | claude-sonnet-4.6 | 25.40 | 11.80 | 34.60 | 1030.00 |
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 1354 individual prompts measured across 1299 monitoring runs over the past month.