Reasoning
CritPt Leaderboard
CritPt (Critical Points) tests graduate-level physics problem-solving on research-style questions spanning mechanics, electromagnetism, quantum mechanics, and more. It was created by physicists to probe whether a model can reason through multi-step physics rather than pattern-match textbook answers.
Source: epoch66 open models ranked+153 proprietaryData through Oct 2026
All models ranked on CritPt
Proprietary / closed models are shown dimmed — you can't run them locally, but they show where the open field stands.
| # | Model | Score | VRAM @ Q4 |
|---|---|---|---|
| 1 | GPT 5.6 Sol Max · proprietary | 32.3% | API only |
| 2 | Claude Opus 5.5 (xhigh) · proprietary | 31.7% | API only |
| 3 | Claude Opus 5.5 Max · proprietary | 31.7% | API only |
| 4 | GPT 6 Astra Max · proprietary | 31.7% | API only |
| 5 | GPT 6.1 Sol (xhigh) · proprietary | 31.7% | API only |
| 6 | GPT 6.1 Sol Max · proprietary | 31.7% | API only |
| 7 | Claude Sonnet 5.5 Max · proprietary | 31.4% | API only |
| 8 | GPT 6 Astra (xhigh) · proprietary | 31.4% | API only |
| 9 | Claude Fable 5.1 (xhigh) · proprietary | 31.1% | API only |
| 10 | Claude Sonnet 5.5 (xhigh) · proprietary | 31.1% | API only |
| 11 | Claude Opus 5.5 (high) · proprietary | 30.9% | API only |
| 12 | GPT 6 Sol Max · proprietary | 30.9% | API only |
| 13 | GPT 5.5 Pro (xhigh) · proprietary | 30.6% | API only |
| 14 | Claude Fable 5.1 (high) · proprietary | 30.3% | API only |
| 15 | GPT 5.4 Pro (Mar 05, 2026, xhigh) · proprietary | 30.0% | API only |
| 16 | GPT 5.6 Terra Max · proprietary | 30.0% | API only |
| 17 | GPT 6.1 Sol (high) · proprietary | 30.0% | API only |
| 18 | Claude Fable 5.1 Max · proprietary | 29.7% | API only |
| 19 | Claude Fable 5.1 (medium) · proprietary | 29.1% | API only |
| 20 | Claude Opus 5 Max · proprietary | 29.1% | API only |
| 21 | GPT 6 Astra (medium) · proprietary | 29.1% | API only |
| 22 | GPT 6 Astra (high) · proprietary | 28.9% | API only |
| 23 | Claude Fable 5 Max · proprietary | 28.6% | API only |
| 24 | GPT 5.6 Sol (xhigh) · proprietary | 28.6% | API only |
| 25 | Claude Opus 5 (high) · proprietary | 28.3% | API only |
| 26 | Muse Wonder 1.0 Max · proprietary | 28.3% | API only |
| 27 | GPT 6 Sol (xhigh) · proprietary | 28.0% | API only |
| 28 | Claude Fable 5.1 (low) · proprietary | 27.7% | API only |
| 29 | Claude Opus 5 (xhigh) · proprietary | 27.7% | API only |
| 30 | Claude Opus 5.5 (medium) · proprietary | 27.7% | API only |
| 31 | GPT 6.1 Sol (medium) · proprietary | 27.7% | API only |
| 32 | Gemini 4 Argon (high) · proprietary | 27.1% | API only |
| 33 | GPT 5.5 (xhigh) · proprietary | 27.1% | API only |
| 34 | GPT 5.6 Terra (xhigh) · proprietary | 27.1% | API only |
| 35 | Claude Opus 5 (medium) · proprietary | 26.9% | API only |
| 36 | MiMo V2.6 Pro RL · 1020B | 26.6% | ~613 GB |
| 37 | GPT 6 Astra (low) · proprietary | 26.3% | API only |
| 38 | Muse Spark 1.3 (xhigh) · proprietary | 26.0% | API only |
| 39 | Gemini 3 Deep Think Preview · proprietary | 25.7% | API only |
| 40 | GPT 5.6 Sol (high) · proprietary | 25.7% | API only |
| 41 | GPT 5.5 (high) · proprietary | 25.4% | API only |
| 42 | GPT 6 Sol (high) · proprietary | 25.4% | API only |
| 43 | GPT 6.1 Sol (low) · proprietary | 24.9% | API only |
| 44 | Muse Spark 1.3 Max · proprietary | 24.9% | API only |
| 45 | Claude Sonnet 5.5 (high) · proprietary | 24.6% | API only |
| 46 | GPT 6 Sol (medium) · proprietary | 24.6% | API only |
| 47 | GPT 5.4 (Mar 05, 2026, xhigh) · proprietary | 23.4% | API only |
| 48 | Kimi K3 · 2779.9B | 23.4% | ~1674 GB |
| 49 | Claude Opus 5 (low) · proprietary | 23.1% | API only |
| 50 | GPT 5.6 Sol (medium) · proprietary | 22.9% | API only |
| 51 | GPT 5.6 Terra (high) · proprietary | 22.9% | API only |
| 52 | Claude Haiku 5.5 (xhigh) · proprietary | 22.6% | API only |
| 53 | Claude Opus 4.8 Max · proprietary | 20.9% | API only |
| 54 | GLM 5.2 · 753.3B | 20.9% | ~456 GB |
| 55 | Step 5 Preview (unspecified) · proprietary | 20.9% | API only |
| 56 | GPT 5.6 Luna (xhigh) · proprietary | 20.6% | API only |
| 57 | GPT 5.6 Luna Max · proprietary | 20.6% | API only |
| 58 | Qwen3.8 Max (unspecified) · proprietary | 20.0% | API only |
| 59 | GPT 6 Astra None · proprietary | 19.7% | API only |
| 60 | Grok 4.6 (xhigh) · proprietary | 19.7% | API only |
| 61 | GPT 6 Luna Max · proprietary | 19.4% | API only |
| 62 | GLM 5.3 · 753.3B | 19.1% | ~456 GB |
| 63 | Claude Haiku 5.5 Max · proprietary | 18.9% | API only |
| 64 | Claude Haiku 5.5 (high) · proprietary | 18.6% | API only |
| 65 | GPT 5.5 (medium) · proprietary | 18.6% | API only |
| 66 | Gemini 3.8 Flash (high) · proprietary | 18.3% | API only |
| 67 | DeepSeek V4 Pro 0813 · 1650.5B | 18.0% | ~991 GB |
| 68 | Grok 4.7 (high) · proprietary | 18.0% | API only |
| 69 | Claude Opus 5.5 (low) · proprietary | 17.7% | API only |
| 70 | Gemini 3.1 Pro Preview · proprietary | 17.7% | API only |
| 71 | Grok 4.6 (medium) · proprietary | 17.7% | API only |
| 72 | Grok 4.7 (xhigh) · proprietary | 17.7% | API only |
| 73 | Muse Spark 1.2 (xhigh) · proprietary | 17.7% | API only |
| 74 | Qwen3.8 Max (Sep 02, unspecified) · proprietary | 17.7% | API only |
| 75 | GPT 5.6 Terra (medium) · proprietary | 17.4% | API only |
| 76 | GPT 6 Luna (xhigh) · proprietary | 17.4% | API only |
| 77 | Grok 4.6 (high) · proprietary | 17.1% | API only |
| 78 | Claude Sonnet 5 Max · proprietary | 16.9% | API only |
| 79 | Claude Sonnet 5.5 (medium) · proprietary | 16.9% | API only |
| 80 | DeepSeek V4 Flash 0731 · 304.2B | 16.6% | ~183 GB |
| 81 | GPT 5.6 Luna (high) · proprietary | 16.6% | API only |
| 82 | GPT 6 Sol (low) · proprietary | 16.3% | API only |
| 83 | Claude Sonnet 5 (xhigh) · proprietary | 15.4% | API only |
| 84 | GLM 5.3 Flash · 321.3B | 15.4% | ~195 GB |
| 85 | GPT 6 Luna (high) · proprietary | 15.4% | API only |
| 86 | Grok 4.5 (high) · proprietary | 15.4% | API only |
| 87 | Muse Spark 1.1 (xhigh) · proprietary | 15.1% | API only |
| 88 | Muse Spark 1.1 · proprietary | 15.1% | API only |
| 89 | GPT 5.6 Sol (low) · proprietary | 14.9% | API only |
| 90 | DeepSeek V4.1 Flash · 763.2B | 14.3% | ~458 GB |
| 91 | Gemini 3.7 Flash (high) · proprietary | 14.3% | API only |
| 92 | Grok 4.7 (low) · proprietary | 13.4% | API only |
| 93 | Qwen3.7 Max · proprietary | 13.4% | API only |
| 94 | Gemini 3.5 Flash (high) · proprietary | 13.1% | API only |
| 95 | Claude Haiku 5.5 (medium) · proprietary | 12.9% | API only |
| 96 | DeepSeek V4 Pro · 1598.8B | 12.9% | ~960 GB |
| 97 | GPT 5 (Aug 07, 2025, high) · proprietary | 12.6% | API only |
| 98 | Gemini 3.8 Flash (medium) · proprietary | 12.3% | API only |
| 99 | Claude Opus 4.7 Max · proprietary | 12.0% | API only |
| 100 | MiMo V2.6 Flash RL · 309B | 12.0% | ~186 GB |
| 101 | Claude Sonnet 5.5 (low) · proprietary | 11.4% | API only |
| 102 | Muse Spark · proprietary | 11.3% | API only |
| 103 | Gemini 3.6 Flash (high) · proprietary | 10.6% | API only |
| 104 | GPT 6 Luna (medium) · proprietary | 10.6% | API only |
| 105 | Mistral Large 4.0 (unspecified) · proprietary | 10.6% | API only |
| 106 | GPT 5.4 Mini (Mar 17, 2026, xhigh) · proprietary | 10.0% | API only |
| 107 | Kimi K2.7 Code · 1026.9B | 10.0% | ~620 GB |
| 108 | Gemini 3.7 Flash (medium) · proprietary | 9.4% | API only |
| 109 | GPT 5.6 Terra (low) · proprietary | 9.4% | API only |
| 110 | GPT 5.4 Nano (Mar 17, 2026, xhigh) · proprietary | 9.3% | API only |
| 111 | Claude Haiku 5.5 (low) · proprietary | 9.1% | API only |
| 112 | Grok Build 0.1 · proprietary | 9.1% | API only |
| 113 | Qwen3.7 Plus · proprietary | 9.1% | API only |
| 114 | Claude Sonnet 5 (medium) · proprietary | 8.6% | API only |
| 115 | Inkling Small · 266.0B | 8.3% | ~160 GB |
| 116 | GPT 5.5 (low) · proprietary | 8.0% | API only |
| 117 | Grok 4.3 (high) · proprietary | 8.0% | API only |
| 118 | Kimi K2.6 · 1026.9B | 8.0% | ~620 GB |
| 119 | DeepSeek V4 Flash · 290.9B | 7.1% | ~175 GB |
| 120 | Gemini 3 Pro Preview · proprietary | 6.9% | API only |
| 121 | Gemini 3.7 Flash (low) · proprietary | 5.7% | API only |
| 122 | Grok 4.6 (low) · proprietary | 5.7% | API only |
| 123 | Inkling · 952.4B | 5.4% | ~572 GB |
| 124 | Qwen3.8 27B · 27.8B | 5.4% | ~17 GB |
| 125 | GPT 5.6 Sol None · proprietary | 5.1% | API only |
| 126 | GPT 5.1 (Nov 13, 2025, unspecified) · proprietary | 4.9% | API only |
| 127 | GPT 5.6 Luna (medium) · proprietary | 4.9% | API only |
| 128 | Claude Sonnet 5 (low) · proprietary | 4.6% | API only |
| 129 | GLM 5.1 · 753.9B | 4.6% | ~457 GB |
| 130 | Gemini 3.8 Flash (low) · proprietary | 4.0% | API only |
| 131 | GPT 6 Sol None · proprietary | 4.0% | API only |
| 132 | MiMo V2.5 Pro · 1023.2B | 4.0% | ~614 GB |
| 133 | MiMo V2.5 · 310.8B | 3.7% | ~187 GB |
| 134 | MiniMax M3 · 427.0B | 3.7% | ~257 GB |
| 135 | Ring 2.6 1T · 1025.4B | 3.7% | ~2057 GB |
| 136 | Claude Sonnet 4.6 Max · proprietary | 3.1% | API only |
| 137 | Kimi K2.5 · 1026.9B | 3.1% | ~620 GB |
| 138 | NVIDIA Nemotron 3 Super 120B A12B BF16 · 123.6B | 3.1% | ~75 GB |
| 139 | NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B | 3.1% | ~370 GB |
| 140 | DeepSeek V3.2 Exp · 685.4B | 2.9% | ~415 GB |
| 141 | Qwen3.6 Plus · proprietary | 2.9% | API only |
| 142 | GPT 5.6 Luna (low) · proprietary | 2.6% | API only |
| 143 | GPT 6 Luna (low) · proprietary | 2.6% | API only |
| 144 | Muse Glimmer 30B · 29.8B | 2.6% | ~18 GB |
| 145 | Step 3.7 Flash · 201.4B | 2.3% | ~133 GB |
| 146 | Gemini 2.5 Pro · proprietary | 2.0% | API only |
| 147 | GPT 5.6 Terra None · proprietary | 2.0% | API only |
| 148 | DeepSeek V3.1 Terminus · 684.5B | 1.7% | ~415 GB |
| 149 | GLM 4.7 · 358.3B | 1.7% | ~216 GB |
| 150 | Gemma 4 31B IT · 31.3B | 1.4% | ~20 GB |
| 151 | GPT 5.5 None · proprietary | 1.4% | API only |
| 152 | GPT OSS 20B · 20.9B | 1.4% | ~13 GB |
| 153 | O3 (Apr 16, 2025, high) · proprietary | 1.4% | API only |
| 154 | Claude Sonnet 4.5 (Sep 29, 2025, unspecified) · proprietary | 1.1% | API only |
| 155 | Claude Sonnet 5 (high) · proprietary | 1.1% | API only |
| 156 | Gemini 3.1 Flash Lite · proprietary | 1.1% | API only |
| 157 | GLM 4.6 · 356.8B | 1.1% | ~215 GB |
| 158 | GPT 6 Luna None · proprietary | 1.1% | API only |
| 159 | GPT OSS 120B · 116.8B | 1.1% | ~71 GB |
| 160 | DeepSeek R1 · 684.5B | 1.1% | ~415 GB |
| 161 | Gemini 2.5 Flash · proprietary | 1.1% | API only |
| 162 | Qwen3.5 122B A10B · 125.1B | 0.9% | ~75 GB |
| 163 | Qwen3.6 27B · 27.8B | 0.9% | ~17 GB |
| 164 | Trinity Large Thinking · 398.6B | 0.9% | ~798 GB |
| 165 | Mercury 2 · proprietary | 0.9% | API only |
| 166 | O4 Mini (Apr 16, 2025, high) · proprietary | 0.6% | API only |
| 167 | MiniMax M2.7 · 228.7B | 0.6% | ~138 GB |
| 168 | Qwen3.5 35B A3B · 36.0B | 0.6% | ~22 GB |
| 169 | Claude Opus 4 (May 14, 2025, unspecified) · proprietary | 0.3% | API only |
| 170 | Claude Sonnet 4 (May 14, 2025, unspecified) · proprietary | 0.3% | API only |
| 171 | Command A Plus 05 2026 BF16 · 218.8B | 0.3% | ~132 GB |
| 172 | GPT 5.6 Luna None · proprietary | 0.3% | API only |
| 173 | Magistral Medium 2509 · proprietary | 0.3% | API only |
| 174 | Magistral Small 2509 · 24.0B | 0.3% | ~15 GB |
| 175 | O3 Mini (Jan 31, 2025, high) · proprietary | 0.3% | API only |
| 176 | Qwen3 30B A3B Thinking 2507 · 30.5B | 0.3% | ~19 GB |
| 177 | Qwen3 32B · 32.8B | 0.3% | ~20 GB |
| 178 | Qwen3.5 9B · 9.7B | 0.3% | ~6.4 GB |
| 179 | Qwen3.6 35B A3B · 36.0B | 0.3% | ~22 GB |
| 180 | Claude 3.5 Haiku (Oct 22, 2024) · proprietary | 0.0% | API only |
| 181 | Claude Haiku 4.5 (Oct 01, 2025, unspecified) · proprietary | 0.0% | API only |
| 182 | Cogito 671B V2.1 · 671B | 0.0% | ~406 GB |
| 183 | DeepSeek v3 · 684.5B | 0.0% | ~415 GB |
| 184 | DeepSeek v3 0324 · 684.5B | 0.0% | ~415 GB |
| 185 | Devstral Small 2 24B Instruct 2512 · 24.0B | 0.0% | ~15 GB |
| 186 | Gemini 3.5 Flash Lite · proprietary | 0.0% | API only |
| 187 | Gemma 3 12B IT · 12.2B | 0.0% | ~8.0 GB |
| 188 | Gemma 3 27B IT · 27.4B | 0.0% | ~18 GB |
| 189 | Gemma 4 26B A4B IT · 25.8B | 0.0% | ~16 GB |
| 190 | GPT 4.1 Mini (Apr 14, 2025) · proprietary | 0.0% | API only |
| 191 | GPT 4.1 Nano (Apr 14, 2025) · proprietary | 0.0% | API only |
| 192 | GPT 4o (Nov 20, 2024) · proprietary | 0.0% | API only |
| 193 | GPT 5 (Aug 07, 2025, minimal) · proprietary | 0.0% | API only |
| 194 | GPT 5 Mini (Aug 07, 2025, high) · proprietary | 0.0% | API only |
| 195 | GPT 5 Mini (Aug 07, 2025, unspecified) · proprietary | 0.0% | API only |
| 196 | GPT 5.5 Instant · proprietary | 0.0% | API only |
| 197 | Granite 4.1 30B · 28.9B | 0.0% | ~18 GB |
| 198 | Grok 4.3 (unspecified) · proprietary | 0.0% | API only |
| 199 | Grok 4.3 None · proprietary | 0.0% | API only |
| 200 | Llama 3.1 8B Instruct · 8.0B | 0.0% | ~5.3 GB |
| 201 | Llama 3.3 70B Instruct · 70.6B | 0.0% | ~47 GB |
| 202 | Llama 4 Maverick 17B 128E Instruct · 401.6B | 0.0% | ~265 GB |
| 203 | Llama 4 Scout 17B 16E Instruct · 108.6B | 0.0% | ~72 GB |
| 204 | Mercury 2.5 (unspecified) · proprietary | 0.0% | API only |
| 205 | MiMo v2 Flash · 309.8B | 0.0% | ~186 GB |
| 206 | Mistral Large 3 675B Instruct 2512 · 675B | 0.0% | ~446 GB |
| 207 | Mistral Medium 2508 · proprietary | 0.0% | API only |
| 208 | Mistral Medium 2604 · proprietary | 0.0% | API only |
| 209 | Mistral Small 3.1 24B Instruct 2503 · 24.0B | 0.0% | ~15 GB |
| 210 | Mistral Small 3.2 24B Instruct 2506 · 24.0B | 0.0% | ~15 GB |
| 211 | Nova 2.0 Pro Preview (low) · proprietary | 0.0% | API only |
| 212 | Nova 2.0 Pro Preview (medium) · proprietary | 0.0% | API only |
| 213 | Nova 2.0 Pro Preview None · proprietary | 0.0% | API only |
| 214 | Phi 4 Mini Instruct · 3.8B | 0.0% | ~2.9 GB |
| 215 | Qwen3 14B · 14.8B | 0.0% | ~9.5 GB |
| 216 | Qwen3 235B A22B Thinking 2507 · 235.1B | 0.0% | ~142 GB |
| 217 | Qwen3 8B · 8.2B | 0.0% | ~5.5 GB |
| 218 | Qwen3 Coder Next · 79.7B | 0.0% | ~48 GB |
| 219 | Solar Pro 3 · proprietary | 0.0% | API only |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Phi 4 Mini Instruct, 4B, score 0.0% — on the efficiency frontier (best score at its size or smaller).
- Qwen3.5 9B, 10B, score 0.3% — on the efficiency frontier (best score at its size or smaller).
- GPT OSS 20B, 21B, score 1.4% — on the efficiency frontier (best score at its size or smaller).
- Qwen3.8 27B, 28B, score 5.4% — on the efficiency frontier (best score at its size or smaller).
- Inkling Small, 266B, score 8.3% — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4 Flash 0731, 304B, score 16.6% — on the efficiency frontier (best score at its size or smaller).
- GLM 5.2, 753B, score 20.9% — on the efficiency frontier (best score at its size or smaller).
- MiMo V2.6 Pro RL, 1T, score 26.6% — on the efficiency frontier (best score at its size or smaller).
CritPt: frequently asked questions
- What is the best open LLM on CritPt?
- MiMo V2.6 Pro RL is the top open model on CritPt, scoring 26.6%. Among all models tested — including proprietary ones — it ranks #36. The top model overall is GPT 5.6 Sol Max (OpenAI) at 32.3%.
- What's the best CritPt model you can run on a 24 GB GPU?
- Qwen3.8 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~17 GB), scoring 5.4% on CritPt.
- What's the best CritPt model you can run on a 12 GB GPU?
- Qwen3.5 9B is the highest-scoring open model that fits in 12 GB at 4-bit quantization (~6.4 GB), scoring 0.3% on CritPt.
- Can open models match proprietary models on CritPt?
- Not quite on CritPt: the strongest proprietary model (GPT 5.6 Sol Max) scores 32.3%, ahead of the best open model (MiMo V2.6 Pro RL) at 26.6% — but you can run the open one yourself.
Scores aggregated from epoch. llmrun does not run this benchmark — see the source for methodology, or the about benchmarks for what it measures.