Rankings
What people actually run
Leaderboards by tokens processed through the gateway, rolled up weekly from the event stream — one immutable event per request, aggregated after the fact.
Top models Sample data
| # | Δ | Model | Tokens | Share | Requests | Apps | In / 1M |
|---|---|---|---|---|---|---|---|
| 1 | — | Claude Sonnet 5anthropic/claude-sonnet-5 | 4.82B | 6.1M | 1240 | $3 | |
| 2 | ▲ 1 | Claude Opus 5anthropic/claude-opus-5 | 3.15B | 2.4M | 980 | $5 | |
| 3 | ▼ 1 | Claude Haiku 4.5anthropic/claude-haiku-4-5 | 2.44B | 9.7M | 860 | $1 | |
| 4 | — | Claude Sonnet 4.6anthropic/claude-sonnet-4-6 | 1.18B | 1.9M | 410 | $3 | |
| 5 | — | Claude Opus 4.8anthropic/claude-opus-4-8 | 742.0M | 610.0K | 320 | $5 | |
| 6 | — | Claude Fable 5anthropic/claude-fable-5 | 486.0M | 210.0K | 145 | $10 | |
| 7 | — | Claude Opus 4.7anthropic/claude-opus-4-7 | 268.0M | 240.0K | 130 | $5 | |
| 8 | — | Claude Opus 4.6anthropic/claude-opus-4-6 | 154.0M | 160.0K | 95 | $5 | |
| 9 | — | Claude Sonnet 4.5anthropic/claude-sonnet-4-5 | 121.0M | 190.0K | 88 | $3 | |
| 10 | — | Claude Opus 4.5anthropic/claude-opus-4-5 | 67.0M | 72.0K | 51 | $5 | |
| 11 | — | Claude Opus 4.1anthropic/claude-opus-4-1 | 21.0M | 26.0K | 19 | $15 |
Trailing 4 weeks Sample data
| # | Model | Tokens | Share | Requests |
|---|---|---|---|---|
| 1 | Claude Sonnet 5anthropic/claude-sonnet-5 | 16.93B | 21.9M | |
| 2 | Claude Haiku 4.5anthropic/claude-haiku-4-5 | 10.42B | 39.3M | |
| 3 | Claude Opus 5anthropic/claude-opus-5 | 10.30B | 7.9M | |
| 4 | Claude Sonnet 4.6anthropic/claude-sonnet-4-6 | 5.45B | 8.7M | |
| 5 | Claude Opus 4.8anthropic/claude-opus-4-8 | 2.79B | 2.3M | |
| 6 | Claude Fable 5anthropic/claude-fable-5 | 1.48B | 640.0K | |
| 7 | Claude Opus 4.7anthropic/claude-opus-4-7 | 1.21B | 1.1M | |
| 8 | Claude Opus 4.6anthropic/claude-opus-4-6 | 698.0M | 724.0K | |
| 9 | Claude Sonnet 4.5anthropic/claude-sonnet-4-5 | 565.0M | 861.0K | |
| 10 | Claude Opus 4.5anthropic/claude-opus-4-5 | 310.0M | 330.0K | |
| 11 | Claude Opus 4.1anthropic/claude-opus-4-1 | 125.0M | 148.0K |
Top models by task Sample data
Coding
Code generation, review, refactors and repository work.
- 1Claude Sonnet 541.798%
- 2Claude Opus 531.91%
- 3Claude Sonnet 4.69.034%
- 4Claude Opus 4.87.146%
- 5Claude Fable 55.933%
Agents
Multi-step tool-using runs and long-horizon autonomy.
- 1Claude Sonnet 538.213%
- 2Claude Opus 533.282%
- 3Claude Haiku 4.59.183%
- 4Claude Opus 4.87.427%
- 5Claude Sonnet 4.66.595%
Extraction
Classification, structured extraction and routing.
- 1Claude Haiku 4.553.185%
- 2Claude Sonnet 528.267%
- 3Claude Sonnet 4.611.228%
- 4Claude Opus 54.202%
- 5Claude Opus 4.82.003%
Writing
Drafting, summarisation and long-form prose.
- 1Claude Sonnet 541.229%
- 2Claude Opus 522.458%
- 3Claude Haiku 4.517.43%
- 4Claude Sonnet 4.611.061%
- 5Claude Opus 4.85.363%
Prices, context windows and capabilities for every model live in thecatalog.