Generalization V2 (Lechmazur): leaderboard
Tests whether LLMs can infer a specific latent theme from limited examples, reject broader but incorrect patterns using anti-examples, and identify the single correct match among eight similar candidates. V2 uses 1,247 validated prompts.
Metric: Inverse-Rank Score. Source: github.com. Saturation forecast: Around January 2027. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 (Thinking, High) | 80.6 | #63 (Claude Opus 4.6) |
| 2 | GPT-5.4 (xHigh) | 80 | #79 (GPT-5.4) |
| 3 | Gemini 3.1 Pro (Preview) | 79.4 | #58 |
| 4 | GPT-5.4 (High) | 77.4 | #79 (GPT-5.4) |
| 5 | GLM-5.1 | 69.8 | #97 |
| 6 | Kimi K2.5 (Thinking) | 69.4 | #144 (Kimi K2.5) |
| 7 | Claude Opus 4.6 (Non-reasoning) | 68.8 | #63 (Claude Opus 4.6) |
| 8 | Claude Sonnet 4.6 (Non-reasoning) | 68.5 | #90 (Claude Sonnet 4.6) |
| 9 | Qwen 3.5 397B A17B | 65.1 | #146 |
| 10 | DeepSeek V3.2 | 65 | #205 |
| 11 | Grok 4.20 0309 (Reasoning) | 63.8 | #112 (Grok 4.20 0309) |
| 12 | Gemini 3.1 Flash Lite (Preview) | 63.3 | #194 |
| 13 | GPT-5.4 Mini (xHigh) | 61.7 | #212 (GPT-5.4 Mini) |
| 14 | Qwen 3.6 Plus | 59.5 | #120 |
| 15 | Seed 2.0 Pro | 57.1 | #118 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=generalization-v2-lechmazur · How It Works · Data refreshed daily, snapshot 2026-10-20.