FiveBench
FiveBench AI study tool leaderboard
FiveBench evaluates how AI models support course building and student learning, with versioned evidence, public artifacts, and uncertainty-aware scoring.
Official FiveBench Index: top five model families
| Rank | Model | Provider | FiveBench Index |
|---|---|---|---|
| 1 | Claude Fable 5 (xhigh) | Anthropic | 70 |
| 2 | GPT-5.6 Sol (xhigh) | OpenAI | 70 |
| 3 | Kimi K3 | Moonshot | 69 |
| 4 | GPT-5.6 Terra (xhigh) | OpenAI | 69 |
| 5 | Qwen 3.8 Max | Qwen | 69 |
Index values are transformed scores, not percentages. Top ordering is provisional where 95% bootstrap intervals overlap.
Methodology
Only complete-profile models are ranked. FiveBench reports raw scores, transformed index values, and confidence intervals; incomplete or retired rows stay available as evidence but are excluded from the official leaderboard.
Current leaderboard build: 2026-08-15. Read the full methodology.