Benchmark
Arena ELO Text Factuality - Math
Math benchmarks measure how well models solve maths problems. Higher scores are better.
All math benchmarksScores from LMArena data.
- Models measured
- 107
- Scores recorded
- 122
- Sources
- 1
- IQ weight
- 2.16
Score distribution
- 1,450.8
- Mean
- 1,446
- Median
- 32.8
- Std dev
- 1,411.7
- 10th percentile
- 1,493.9
- 90th percentile
- 1,358
- Min
This benchmark has a weight of 2.16 for chat models, measured over 107 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 101 | MiniMax-M2.1 Preview· LMArena data | 1,422 |
| 102 | Step 3.5 Flash· LMArena data | 1,420.3 |
| 103 | Mistral Large 3· LMArena data | 1,420.1 |
| 104 | Doubao Seed 2.0 Pro· LMArena data | 1,417.5 |
| 105 | ChatGPT-4o Latest20250326· LMArena data | 1,417.3 |
| 106 | Mistral Medium 3.12508· LMArena data | 1,414.1 |
| 107 | Gemini 2.5 Flash· LMArena data | 1,413.7 |
| 108 | Gemini 2.5 Flash Preview09-2025· LMArena data | 1,413.6 |
| 109 | Claude Haiku 4.520251001· LMArena data | 1,413.6 |
| 110 | Grok 4.1 FastReasoning· LMArena data | 1,411.7 |
| 111 | GPT-5.3 Chat· LMArena data | 1,411.4 |
| 112 | Qwen3.5 35B A3B· LMArena data | 1,408.1 |
| 113 | Qwen3.5 Flash· LMArena data | 1,406.3 |
| 114 | Grok 4.3· LMArena data | 1,406.3 |
| 115 | Grok 4.1· LMArena data | 1,405.6 |
| 116 | MiMo V2 FlashNon-Thinking· LMArena data | 1,402.1 |
| 117 | Gemini 2.5 Flash-Lite PreviewNon-Thinking, 09-2025· LMArena data | 1,393 |
| 118 | Nemotron 3.5 Lightning 30B A3BNVFP4· LMArena data | 1,392.4 |
| 119 | Trinity Large Thinking· LMArena data | 1,383.9 |
| 120 | Nemotron 3 Nano 30B A3BBF16· LMArena data | 1,380.4 |
| 121 | MiniMax M2.5· LMArena data | 1,378 |
| 122 | Trinity Large Preview· LMArena data | 1,358 |
Showing 101–122 of 122 results.
Data credits: LMArena, CC BY 4.0 (licence)