isitsota / categories / Math
Reasoning / knowledge · Oct 13, 2026
Who is SOTA in Math?
GPT-6 Astra & Gemini 3.1 Pro Preview & Gemini 3.8 Flash & Kimi K2 5 & GPT-5 Mini & Gemini 3.1 Flash Lite & GLM 4.5 Air
Research-math and contest-math evidence. FrontierMath is preferred over saturated high-school sets. As of Oct 13, 2026, GPT-6 Astra & Gemini 3.1 Pro Preview & Gemini 3.8 Flash & Kimi K2 5 & GPT-5 Mini & Gemini 3.1 Flash Lite & GLM 4.5 Air holds Math. On…
Standings · pareto
The board
| # | Model | State | Confidence |
|---|---|---|---|
| 1 | OGPT-6 Astra OpenAI | Pareto | medium |
| 2 | AClaude Fable 5.1 Anthropic | Near-SOTA | medium |
| 3 | GGemini 3.1 Pro Preview Google DeepMind | Pareto | medium |
| 4 | GGemini 3.8 Flash Google DeepMind | Pareto | medium |
| 5 | OGPT-5.4 Pro OpenAI | Not SOTA · #5 | medium |
| 8 | OGPT-5.4 OpenAI | Not SOTA · #8 | medium |
| 9 | AClaude Opus 4.7 Anthropic | Not SOTA · #9 | medium |
| 10 | AClaude Opus 4.6 Anthropic | Not SOTA · #10 | medium |
| 11 | OGPT-5 Pro OpenAI | Not SOTA · #11 | medium |
| 12 | OGPT-5.2 OpenAI | Not SOTA · #12 | medium |
| 13 | OGPT-5 OpenAI | Not SOTA · #13 | medium |
| 14 | AClaude Opus 4.5 Anthropic | Not SOTA · #14 | medium |
| 15 | KKimi K2 5 Moonshot AI | Pareto | medium |
| 16 | OGPT-5.1 OpenAI | Not SOTA · #16 | medium |
| 18 | OO3 OpenAI | Not SOTA · #18 | medium |
Pareto
Quality against blended price
Evidence recipe
What can move this crown
- Humanity's Last Exam — primary, weight 1
- FrontierMath — primary, weight 1 (display only, cannot crown)
- LMArena Text (style-controlled, math) — primary, weight 0.8 (display only, cannot crown)