← Global
instruction following
instruction following ranking
index v0.1.0 · run cmsfulul · 2026-08-05 08:51 UTC| # | Model | Provider | Score (95% CI) | Consist. | Calib. | Contam. Δ | p50 | $/1k | |
|---|---|---|---|---|---|---|---|---|---|
| 1 |  Amazon: Nova Premier 1.0 | amazon | 757 [600–913] | 0.98 | 1.00 | 0.000 | 296ms | $2.11 | |
| 2 |  Anthropic: Claude Sonnet 4.6 | anthropic | 714 [565–864] | 1.00 | 0.98 | 0.038 | 1.6s | $1.17 | |
| 3 | Anthropic: Claude Sonnet 5 | anthropic | 711 [562–861] | 0.98 | 0.97 | 0.038 | 2.6s | $1.05 | |
| 4 |  DeepSeek: DeepSeek V4 Flash 0731 | deepseek | 710 [561–860] | 0.98 | 0.97 | 0.038 | 854ms | $0.031 | |
| 5 | Anthropic: Claude Fable 5 | anthropic | 627 [529–725] | 1.00 | 0.98 | 0.000 | 3.7s | $5.62 |
- 757 [600–913]consist 0.98calib 1.00Δ 0.000296ms$2.11/1k
- #2Anthropic: Claude Sonnet 4.6anthropic714 [565–864]consist 1.00calib 0.98Δ 0.0381.6s$1.17/1k
- #3Anthropic: Claude Sonnet 5anthropic711 [562–861]consist 0.98calib 0.97Δ 0.0382.6s$1.05/1k
- 710 [561–860]consist 0.98calib 0.97Δ 0.038854ms$0.031/1k
- #5Anthropic: Claude Fable 5anthropic627 [529–725]consist 1.00calib 0.98Δ 0.0003.7s$5.62/1k