Benchmarks · reasoning
TasteVal
Uses experimental-compute efficiency relative to human experts, divided by the current highest model multiplier. Retains the native multiplier and 95% source interval. Score: compute multiplier / current leaderboard maximum. Collected by Jake Boggs' benchmark dashboard.
20 models · 20 results · data as of 2026-10-29 · pzeroresearch.com ↗
Leaderboard
Opus 5.5 max
All results
20 rows| # | Model | Effort | TasteVal |
|---|---|---|---|
| 1 | Opus 5.5 | max | 100.0% |
| 2 | Fable 5.1 | max | 74.5% |
| 3 | GPT-6 Astra | max | 69.5% |
| 4 | Opus 5 | max | 55.8% |
| 5 | Opus 4.8 | max | 36.5% |
| 6 | GLM 5.3 | max | 25.7% |
| 7 | GPT-5.6 Sol | max | 25.6% |
| 8 | Opus 4.7 | max | 21.7% |
| 9 | GPT-5.5 | xhigh | 20.4% |
| 10 | Kimi K3 | max | 19.7% |
| 11 | Opus 4.6 | max | 17.5% |
| 12 | GPT-5.2 | high | 12.0% |
| 13 | GPT-5.4 | xhigh | 9.2% |
| 14 | Opus 4.5 | high | 5.1% |
| 15 | GPT-5 | high | 5.0% |
| 16 | Kimi K2.5 | max | 3.7% |
| 17 | o3 | high | 3.6% |
| 18 | o1 | high | 2.9% |
| 19 | GPT-4o | high | 2.6% |
| 20 | GPT-4 | high | 1.2% |