Automated page. Written and translated automatically by AI and checked against the sources by software; no human editor. SI-IQ is an index built from public benchmark results, not an IQ test. How it works
The benchmarks behind SI-IQ
SI-IQ currently draws on 14 benchmarks. Each is run by an independent evaluator; developers' own claims are never used in the index.
| Benchmark | Area | Evaluated by | Models |
|---|---|---|---|
| Chess Puzzles | Reasoning | Epoch AI | 81 |
| Furniture Assembly | Reasoning | Epoch AI | 34 |
| Mystery Game Puzzles | Reasoning | Epoch AI | 78 |
| FrontierMath Tier 4 | Math | Epoch AI | 68 |
| FrontierMath Tiers 1–3 | Math | Epoch AI | 86 |
| MATH Level 5 | Math | Epoch AI | 86 |
| OTIS Mock AIME 2024–2025 | Math | Epoch AI | 193 |
| Aider Polyglot | Coding | Aider | 40 |
| SWE-bench Verified (Epoch AI run) | Coding | Epoch AI | 32 |
| GPQA Diamond | Knowledge | Epoch AI | 205 |
| GPQA Diamond (OpenRouter run) | Knowledge | OpenRouter | 117 |
| SimpleQA Verified | Knowledge | Epoch AI | 81 |
| Berkeley Function Calling Leaderboard | Agents | UC Berkeley (Gorilla) | 41 |
| τ²-bench Airline (OpenRouter run) | Agents | OpenRouter | 117 |
Loading k shows how strongly a benchmark separates stronger from weaker models on the common ability scale (higher = more informative). Read the full methodology
Data sources and licences
| Source | Licence / terms | In our download |
|---|---|---|
| Epoch AI Benchmarking Hub | CC BY 4.0 | Yes |
| OpenRouter benchmark datasets | CC BY 4.0 | Yes |
| Aider LLM Leaderboards | Apache-2.0 | Yes |
| Berkeley Function Calling Leaderboard | Apache-2.0 | Yes |
| Arena (formerly LMArena) leaderboard dataset | CC BY 4.0 | Yes |
| LiteLLM model price table | MIT | Yes |