Reasoning
GPQA Diamond Leaderboard
GPQA Diamond is a set of extremely hard, graduate-level science questions (physics, chemistry, biology) written by domain experts and filtered so that skilled non-experts with web access still fail. It measures genuine reasoning rather than memorization.
Source: epoch98 open models ranked+202 proprietaryData through Oct 2026
All models ranked on GPQA Diamond
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 | Gemini 4 Argon (high) · proprietary | 96.4% | API only |
| 2 | GPT 6 Astra Max · proprietary | 95.8% | API only |
| 3 | Claude Sonnet 5.5 Max · proprietary | 95.6% | API only |
| 4 | Gemini 3.8 Flash (high) · proprietary | 95.4% | API only |
| 5 | GPT 6.1 Sol Max · proprietary | 95.4% | API only |
| 6 | Gemini 3.7 Flash (high) · proprietary | 94.8% | API only |
| 7 | GPT 5.4 Pro (Mar 05, 2026, xhigh) · proprietary | 94.6% | API only |
| 8 | Gemini 3.1 Pro Preview (high) · proprietary | 94.4% | API only |
| 9 | GPT 6 Sol Max · proprietary | 94.3% | API only |
| 10 | Gemini 3.6 Flash (high) · proprietary | 94.1% | API only |
| 11 | Gemini 3.1 Pro Preview · proprietary | 94.1% | API only |
| 12 | GPT 5.5 Pre Release (xhigh) · proprietary | 94.0% | API only |
| 13 | Grok 4.6 (high) · proprietary | 94.0% | API only |
| 14 | GPT 5.5 Pro Pre Release (xhigh) · proprietary | 93.9% | API only |
| 15 | Claude Opus 5 Max · proprietary | 93.9% | API only |
| 16 | GPT 5.6 Sol Max · proprietary | 93.5% | API only |
| 17 | Grok 4.5 (high) · proprietary | 93.4% | API only |
| 18 | GPT 5.6 Terra Max · proprietary | 93.3% | API only |
| 19 | GPT 5.4 (Mar 05, 2026, xhigh) · proprietary | 93.3% | API only |
| 20 | Grok 4.6 (xhigh) · proprietary | 93.2% | API only |
| 21 | Kimi K3 · 2779.9B | 93.1% | ~1674 GB |
| 22 | Claude Opus 5 · proprietary | 92.9% | API only |
| 23 | Gemini 3.5 Flash (high) · proprietary | 92.8% | API only |
| 24 | Grok 4.7 (xhigh) · proprietary | 92.7% | API only |
| 25 | Qwen3.8 Max (xhigh) · proprietary | 92.7% | API only |
| 26 | Gemini 3 Pro Preview · proprietary | 92.6% | API only |
| 27 | Qwen3.8 Max (Sep 02, xhigh) · proprietary | 92.3% | API only |
| 28 | GLM 5.2 · 753.3B | 91.9% | ~456 GB |
| 29 | DeepSeek V4 Pro 0813 · 1650.5B | 91.7% | ~991 GB |
| 30 | GPT 5.6 Luna Max · proprietary | 91.6% | API only |
| 31 | GPT 5.2 (Dec 11, 2025, xhigh) · proprietary | 91.4% | API only |
| 32 | Claude Opus 4.8 Max · proprietary | 91.0% | API only |
| 33 | DeepSeek V4 Flash 0731 · 304.2B | 91.0% | ~183 GB |
| 34 | DeepSeek V4 Pro · 1598.8B | 90.9% | ~960 GB |
| 35 | GLM 5.3 · 753.3B | 90.9% | ~456 GB |
| 36 | MiniMax M3 · 427.0B | 90.9% | ~257 GB |
| 37 | Qwen3.7 Max · proprietary | 90.9% | API only |
| 38 | Kimi K2.6 · 1026.9B | 90.8% | ~620 GB |
| 39 | GPT 5.5 (low) · proprietary | 90.7% | API only |
| 40 | Claude Opus 5.5 Max · proprietary | 90.6% | API only |
| 41 | Claude Opus 4.6 (32K) · proprietary | 90.5% | API only |
| 42 | Claude Sonnet 5 (xhigh) · proprietary | 90.5% | API only |
| 43 | GPT 6 Luna Max · proprietary | 90.5% | API only |
| 44 | Claude Opus 4.7 (xhigh) · proprietary | 90.1% | API only |
| 45 | GLM 5.3 Flash · 321.3B | 90.1% | ~195 GB |
| 46 | GLM 5.1 · 753.9B | 89.9% | ~457 GB |
| 47 | GPT 5.4 (Mar 05, 2026, high) · proprietary | 89.9% | API only |
| 48 | GPT 5.6 Sol (low) · proprietary | 89.9% | API only |
| 49 | DeepSeek V4.1 Flash · 763.2B | 89.8% | ~458 GB |
| 50 | Muse Spark · proprietary | 89.8% | API only |
| 51 | Claude Haiku 5.5 Max · proprietary | 89.6% | API only |
| 52 | Gemini 3 Flash Preview (high) · proprietary | 89.4% | API only |
| 53 | Grok 4.20 0309 Reasoning · proprietary | 89.3% | API only |
| 54 | Gemini 3.5 Flash (low) · proprietary | 88.9% | API only |
| 55 | GPT 5.4 (Mar 05, 2026, medium) · proprietary | 88.9% | API only |
| 56 | Grok 4.3 (high) · proprietary | 88.8% | API only |
| 57 | Claude Opus 4.6 (64K) · proprietary | 88.8% | API only |
| 58 | Muse Spark Open 816B A42B · 822.0B | 88.6% | ~543 GB |
| 59 | Inkling Small · 266.0B | 88.5% | ~160 GB |
| 60 | Claude Opus 4.6 Max · proprietary | 88.4% | API only |
| 61 | Claude Opus 4.8 (low) · proprietary | 88.4% | API only |
| 62 | Qwen3.6 Plus · proprietary | 88.4% | API only |
| 63 | Inkling · 952.4B | 88.3% | ~572 GB |
| 64 | GPT 5.2 (Dec 11, 2025, high) · proprietary | 88.2% | API only |
| 65 | Mistral Large 4.0 (high) · proprietary | 88.1% | API only |
| 66 | Claude Opus 5 (low) · proprietary | 87.9% | API only |
| 67 | GPT 5.2 (Dec 11, 2025, medium) · proprietary | 87.9% | API only |
| 68 | Kimi K2.7 Code · 1026.9B | 87.9% | ~620 GB |
| 69 | Qwen3.7 Plus · proprietary | 87.9% | API only |
| 70 | GLM 5 · 753.9B | 87.8% | ~457 GB |
| 71 | GPT 5.1 (Nov 13, 2025, high) · proprietary | 87.6% | API only |
| 72 | Kimi K2.5 · 1026.9B | 87.6% | ~620 GB |
| 73 | Claude Sonnet 4.6 (32K) · proprietary | 87.4% | API only |
| 74 | GPT 5.6 Terra (low) · proprietary | 87.4% | API only |
| 75 | Qwen3.6 Max Preview · proprietary | 87.4% | API only |
| 76 | Grok 4 (Jul 09) · proprietary | 87.0% | API only |
| 77 | GPT 5.4 Mini (Mar 17, 2026, xhigh) · proprietary | 86.9% | API only |
| 78 | Claude Opus 4.7 Max · proprietary | 86.4% | API only |
| 79 | Gemini 3.5 Flash (minimal) · proprietary | 86.4% | API only |
| 80 | Gemini 3.6 Flash (low) · proprietary | 86.4% | API only |
| 81 | Qwen3.5 397B A17B · 403.4B | 86.4% | ~242 GB |
| 82 | GPT 5 (Aug 07, 2025, high) · proprietary | 86.2% | API only |
| 83 | Claude Opus 4.5 (Nov 01, 2025, 32K) · proprietary | 86.1% | API only |
| 84 | Claude Fable 5 Max · proprietary | 85.9% | API only |
| 85 | Gemini 3.6 Flash (minimal) · proprietary | 85.9% | API only |
| 86 | Qwen3.6 27B · 27.8B | 85.9% | ~17 GB |
| 87 | Claude Opus 4.5 (Nov 01, 2025, 16K) · proprietary | 85.5% | API only |
| 88 | Claude Opus 4.8 None · proprietary | 85.4% | API only |
| 89 | GPT 5 (Aug 07, 2025, medium) · proprietary | 85.4% | API only |
| 90 | NVIDIA Nemotron 3 Ultra 550B A55B BF16 · 560.5B | 85.4% | ~370 GB |
| 91 | Gemini 2.5 Pro · proprietary | 85.3% | API only |
| 92 | GPT 5.1 (Nov 13, 2025, medium) · proprietary | 85.0% | API only |
| 93 | Gemini 2.5 Pro Preview (Jun 05) · proprietary | 84.9% | API only |
| 94 | GPT 5.4 (Mar 05, 2026, low) · proprietary | 84.9% | API only |
| 95 | Qwen3.5 Plus · proprietary | 84.9% | API only |
| 96 | Qwen3.6 35B A3B · 36.0B | 84.9% | ~22 GB |
| 97 | Kimi K2 Thinking · 1026.4B | 84.2% | ~620 GB |
| 98 | Gemini 2.5 Pro Exp (Mar 25) · proprietary | 83.8% | API only |
| 99 | GPT 5.4 Mini (Mar 17, 2026, high) · proprietary | 83.6% | API only |
| 100 | Qwen3.5 35B A3B · 36.0B | 83.5% | ~22 GB |
| 101 | DeepSeek Reasoner · proprietary | 83.4% | API only |
| 102 | Claude Fable 5 (high) · proprietary | 83.3% | API only |
| 103 | Claude Sonnet 4.6 (high) · proprietary | 83.3% | API only |
| 104 | Claude Sonnet 4.6 (medium) · proprietary | 83.3% | API only |
| 105 | Gemini 3.5 Flash Lite (high) · proprietary | 83.3% | API only |
| 106 | GLM 4.7 · 358.3B | 83.3% | ~216 GB |
| 107 | Qwen3.6 Flash · proprietary | 83.3% | API only |
| 108 | Gemini 3 Flash Preview · proprietary | 83.2% | API only |
| 109 | GPT 5.6 Sol None · proprietary | 82.8% | API only |
| 110 | GPT 5.2 (Dec 11, 2025, low) · proprietary | 82.7% | API only |
| 111 | GPT 5.5 Instant · proprietary | 82.5% | API only |
| 112 | Claude Sonnet 4.5 (Sep 29, 2025, 59K) · proprietary | 82.3% | API only |
| 113 | GPT 5.6 Luna (low) · proprietary | 82.3% | API only |
| 114 | Qwen3.5 Flash · proprietary | 82.3% | API only |
| 115 | Qwen3.7 Flash · proprietary | 82.3% | API only |
| 116 | Gemini 3.1 Flash Lite (high) · proprietary | 81.8% | API only |
| 117 | O3 (Apr 16, 2025, high) · proprietary | 81.8% | API only |
| 118 | Qwen3.7 Plus None · proprietary | 81.8% | API only |
| 119 | Claude Sonnet 4.5 (Sep 29, 2025, 32K) · proprietary | 81.7% | API only |
| 120 | O3 (Apr 16, 2025, medium) · proprietary | 80.8% | API only |
| 121 | Qwen3.7 Flash None · proprietary | 80.8% | API only |
| 122 | Claude Opus 4.5 (Nov 01, 2025) · proprietary | 80.7% | API only |
| 123 | Claude Sonnet 5 Max · proprietary | 80.3% | API only |
| 124 | Qwen3 235B A22B Thinking 2507 · 235.1B | 80.0% | ~142 GB |
| 125 | O3 (Apr 16, 2025, low) · proprietary | 79.8% | API only |
| 126 | O4 Mini (Apr 16, 2025, high) · proprietary | 79.6% | API only |
| 127 | Qwen3.5 9B · 9.7B | 79.0% | ~6.4 GB |
| 128 | Claude Fable 5 (low) · proprietary | 78.8% | API only |
| 129 | Claude Sonnet 4.5 (Sep 29, 2025, 16K) · proprietary | 78.8% | API only |
| 130 | Claude Sonnet 4.6 Max · proprietary | 78.8% | API only |
| 131 | Claude 3.7 Sonnet (Feb 19, 2025, 64K) · proprietary | 78.5% | API only |
| 132 | GPT 5.4 Nano (Mar 17, 2026, high) · proprietary | 78.5% | API only |
| 133 | Claude Sonnet 4 (May 14, 2025, 32K) · proprietary | 78.3% | API only |
| 134 | Claude Sonnet 4 (May 14, 2025, 59K) · proprietary | 77.8% | API only |
| 135 | O4 Mini (Apr 16, 2025, medium) · proprietary | 77.8% | API only |
| 136 | Claude Opus 4.1 (Aug 05, 2025, 16K) · proprietary | 77.3% | API only |
| 137 | GPT 5.5 None · proprietary | 77.3% | API only |
| 138 | GPT 5.6 Terra None · proprietary | 77.3% | API only |
| 139 | O3 Mini (Jan 31, 2025, high) · proprietary | 77.0% | API only |
| 140 | Claude 3.7 Sonnet (Feb 19, 2025, 16K) · proprietary | 76.8% | API only |
| 141 | Claude 3.7 Sonnet (Feb 19, 2025, 32K) · proprietary | 76.8% | API only |
| 142 | Claude Opus 4.1 (Aug 05, 2025, 27K) · proprietary | 76.8% | API only |
| 143 | O1 (Dec 17, 2024, high) · proprietary | 76.8% | API only |
| 144 | DeepSeek R1 0528 · 684.5B | 76.3% | ~415 GB |
| 145 | Claude Opus 4 (May 14, 2025, 16K) · proprietary | 76.3% | API only |
| 146 | Grok 3 Mini Beta (low) · proprietary | 76.3% | API only |
| 147 | Claude Sonnet 4 (May 14, 2025, 16K) · proprietary | 75.8% | API only |
| 148 | Gemini 3.5 Flash Lite (low) · proprietary | 75.8% | API only |
| 149 | Gemma 4 31B IT · 31.3B | 75.8% | ~20 GB |
| 150 | GPT OSS 120B · 116.8B | 75.8% | ~71 GB |
| 151 | O1 (Dec 17, 2024, medium) · proprietary | 75.8% | API only |
| 152 | O4 Mini (Apr 16, 2025, low) · proprietary | 75.3% | API only |
| 153 | GPT 5 Mini (Aug 07, 2025, high) · proprietary | 75.0% | API only |
| 154 | GPT 5.4 2026 03.05 None · proprietary | 74.8% | API only |
| 155 | Grok 3 Mini Beta (high) · proprietary | 74.6% | API only |
| 156 | O3 Mini (Jan 31, 2025, medium) · proprietary | 74.3% | API only |
| 157 | Gemini 3.1 Flash Lite (low) · proprietary | 74.2% | API only |
| 158 | Gemini 3.5 Flash Lite (minimal) · proprietary | 74.2% | API only |
| 159 | O1 (Dec 17, 2024, low) · proprietary | 74.2% | API only |
| 160 | Claude Sonnet 4.5 (Sep 29, 2025) · proprietary | 73.7% | API only |
| 161 | Gemini 3.1 Flash Lite (minimal) · proprietary | 73.7% | API only |
| 162 | Claude Opus 4.1 (Aug 05, 2025) · proprietary | 73.2% | API only |
| 163 | Gemma 4 26B A4B IT · 25.8B | 73.2% | ~16 GB |
| 164 | GPT 5.2 2025 12.11 None · proprietary | 73.2% | API only |
| 165 | Qwen3 Max (Sep 23, 2025) · proprietary | 72.6% | API only |
| 166 | GPT 5.4 Nano (Mar 17, 2026, low) · proprietary | 72.2% | API only |
| 167 | GPT 5 (Aug 07, 2025, minimal) · proprietary | 71.7% | API only |
| 168 | GPT 5 Mini (Aug 07, 2025, minimal) · proprietary | 71.7% | API only |
| 169 | GPT 5 Mini (Aug 07, 2025, medium) · proprietary | 71.7% | API only |
| 170 | Seed OSS 36B Instruct · 36.2B | 71.5% | ~22 GB |
| 171 | Claude Haiku 4.5 (Oct 01, 2025, 32K) · proprietary | 71.2% | API only |
| 172 | DeepSeek Chat · proprietary | 71.2% | API only |
| 173 | Qwen3 235B A22B · 235.1B | 70.7% | ~142 GB |
| 174 | Qwen3 30B A3B Thinking 2507 · 30.5B | 70.1% | ~19 GB |
| 175 | GPT 5 Nano (Aug 07, 2025, high) · proprietary | 69.4% | API only |
| 176 | DeepSeek R1 · 684.5B | 69.2% | ~415 GB |
| 177 | Claude Opus 4 (May 14, 2025) · proprietary | 69.2% | API only |
| 178 | GPT 4.5 Preview (Feb 27, 2025) · proprietary | 68.7% | API only |
| 179 | O3 Mini (Jan 31, 2025, low) · proprietary | 68.2% | API only |
| 180 | DeepSeek v3 0324 · 684.5B | 67.6% | ~415 GB |
| 181 | Grok 3 Beta · proprietary | 67.6% | API only |
| 182 | GPT 5 Nano (Aug 07, 2025, medium) · proprietary | 67.4% | API only |
| 183 | Llama 4 Maverick 17B 128E Instruct · 401.6B | 67.0% | ~265 GB |
| 184 | GPT 4.1 (Apr 14, 2025) · proprietary | 66.9% | API only |
| 185 | Claude Sonnet 4 (May 14, 2025) · proprietary | 66.7% | API only |
| 186 | Gemini 2.5 Pro Preview (May 06) · proprietary | 66.7% | API only |
| 187 | GPT 5.1 2025 11.13 None · proprietary | 66.7% | API only |
| 188 | Claude 3.7 Sonnet (Feb 19, 2025) · proprietary | 66.0% | API only |
| 189 | GPT 4.1 Mini (Apr 14, 2025) · proprietary | 65.8% | API only |
| 190 | Qwen3 32B · 32.8B | 65.7% | ~20 GB |
| 191 | Gemini 2.0 Pro Exp (Feb 05) · proprietary | 65.7% | API only |
| 192 | QwQ Plus · proprietary | 65.4% | API only |
| 193 | QwQ 32B · 32.8B | 65.3% | ~21 GB |
| 194 | DeepSeek R1 Distill Qwen 32B · 32.8B | 64.1% | ~21 GB |
| 195 | Gemini 2.0 Flash 001 · proprietary | 64.1% | API only |
| 196 | GPT 5.4 Mini 2026 03.17 None · proprietary | 64.1% | API only |
| 197 | Qwen3 14B · 14.8B | 63.8% | ~9.5 GB |
| 198 | GPT 5.6 Luna None · proprietary | 63.6% | API only |
| 199 | O1 Mini (Sep 12, 2024, high) · proprietary | 62.4% | API only |
| 200 | Qwen3 30B A3B · 30.5B | 61.7% | ~19 GB |
| 201 | GPT OSS 20B · 20.9B | 60.8% | ~13 GB |
| 202 | GLM 4.7 Flash · 31.2B | 60.5% | ~20 GB |
| 203 | Claude Haiku 4.5 (Oct 01, 2025) · proprietary | 60.5% | API only |
| 204 | Mistral Medium 2505 · proprietary | 59.5% | API only |
| 205 | O1 Mini (Sep 12, 2024, medium) · proprietary | 59.5% | API only |
| 206 | GPT 5 Nano (Aug 07, 2025, low) · proprietary | 57.6% | API only |
| 207 | Gemini 1.5 Pro 002 · proprietary | 57.2% | API only |
| 208 | Gemini 2.0 Flash Thinking Exp (Jan 21) · proprietary | 57.1% | API only |
| 209 | Qwen3 8B · 8.2B | 56.8% | ~5.5 GB |
| 210 | DeepSeek v3 · 684.5B | 56.5% | ~415 GB |
| 211 | Qwen Max (Jan 25, 2025) · proprietary | 56.1% | API only |
| 212 | Magistral Small 2506 · 23.6B | 56.1% | ~15 GB |
| 213 | Phi 4 · 14.7B | 56.1% | ~9.5 GB |
| 214 | DeepSeek R1 Distill Llama 70B · 70.6B | 55.7% | ~43 GB |
| 215 | Qwen3 30B A3B Instruct 2507 · 30.5B | 55.6% | ~19 GB |
| 216 | GPT 5.4 Nano 2026 03.17 None · proprietary | 55.6% | API only |
| 217 | Claude 3.5 Sonnet (Oct 22, 2024) · proprietary | 55.3% | API only |
| 218 | Claude 3.5 Sonnet (Jun 20, 2024) · proprietary | 54.0% | API only |
| 219 | Grok 2 (Dec 12) · proprietary | 53.8% | API only |
| 220 | Qwen3 4B · 4.0B | 52.3% | ~2.9 GB |
| 221 | Llama 4 Scout 17B 16E Instruct · 108.6B | 51.8% | ~72 GB |
| 222 | Mistral Large Instruct 2411 · 122.6B | 51.3% | ~75 GB |
| 223 | Llama 3.1 405B Instruct · 405.9B | 50.9% | ~268 GB |
| 224 | O1 Preview (Sep 12, 2024) · proprietary | 50.3% | API only |
| 225 | GPT 4o (Aug 06, 2024) · proprietary | 49.2% | API only |
| 226 | Qwen2.5 72B Instruct · 72.7B | 49.1% | ~45 GB |
| 227 | Mistral Small 3.2 24B Instruct 2506 · 24.0B | 49.0% | ~15 GB |
| 228 | Mistral Large Instruct 2407 · 122.6B | 49.0% | ~81 GB |
| 229 | GPT 4.1 Nano (Apr 14, 2025) · proprietary | 48.9% | API only |
| 230 | GPT 4o (May 13, 2024) · proprietary | 48.9% | API only |
| 231 | GPT 5 Nano (Aug 07, 2025, minimal) · proprietary | 48.5% | API only |
| 232 | Qwen Plus (Jan 25, 2025) · proprietary | 48.1% | API only |
| 233 | GPT 4o (Nov 20, 2024) · proprietary | 47.9% | API only |
| 234 | Gemma 3 27B IT · 27.4B | 47.7% | ~18 GB |
| 235 | Magistral Small 2509 · 24.0B | 47.6% | ~15 GB |
| 236 | Mistral Small 3.1 24B Instruct 2503 · 24.0B | 47.5% | ~15 GB |
| 237 | Llama 3.3 70B Instruct · 70.6B | 47.4% | ~47 GB |
| 238 | Gemini 1.5 Flash 002 · proprietary | 47.3% | API only |
| 239 | Mistral Small 24B Instruct 2501 · 23.6B | 47.3% | ~15 GB |
| 240 | Claude 3 Opus (Feb 29, 2024) · proprietary | 47.2% | API only |
| 241 | GPT 4 Turbo (Apr 09, 2024) · proprietary | 46.6% | API only |
| 242 | Llama 3.1 Tulu 3 70B DPO · 70.6B | 46.3% | ~43 GB |
| 243 | Qwen2.5 32B Instruct · 32.8B | 46.1% | ~21 GB |
| 244 | Gemini 1.5 Pro 001 · proprietary | 45.9% | API only |
| 245 | Qwen3 4B Instruct 2507 · 4.0B | 45.8% | ~2.9 GB |
| 246 | DeepSeek R1 Distill Qwen 14B · 14.8B | 44.7% | ~9.6 GB |
| 247 | Llama 3.1 70B Instruct · 70.6B | 44.2% | ~47 GB |
| 248 | WizardLM 2 8x22B · 140.6B | 43.4% | ~85 GB |
| 249 | GPT 4 1106 Preview · proprietary | 42.4% | API only |
| 250 | GPT 4 0125 Preview · proprietary | 42.3% | API only |
| 251 | Qwen Turbo (Nov 01, 2024) · proprietary | 41.8% | API only |
| 252 | Llama 3.2 90B Vision Instruct · 88.6B | 41.0% | ~195 GB |
| 253 | Qwen2 72B Instruct · 72.7B | 40.8% | ~45 GB |
| 254 | Claude 3 Sonnet (Feb 29, 2024) · proprietary | 40.6% | API only |
| 255 | Meta Llama 3 70B Instruct · 70.6B | 40.6% | ~47 GB |
| 256 | Gemini 1.5 Flash 001 · proprietary | 40.4% | API only |
| 257 | Gemma 3 12B IT · 12.2B | 39.5% | ~8.0 GB |
| 258 | Mistral Large 2402 · proprietary | 38.8% | API only |
| 259 | Claude 3.5 Haiku (Oct 22, 2024) · proprietary | 38.1% | API only |
| 260 | Qwen3 1.7B · 2.0B | 38.0% | ~1.8 GB |
| 261 | GPT 4o Mini (Jul 18, 2024) · proprietary | 37.7% | API only |
| 262 | Hermes 2 Theta Llama 3 70B · 70.6B | 37.5% | ~43 GB |
| 263 | Gemma 2 27B IT · 27.2B | 36.5% | ~18 GB |
| 264 | Claude 3 Haiku (Mar 07, 2024) · proprietary | 36.3% | API only |
| 265 | GPT 4 (Mar 14) · proprietary | 35.7% | API only |
| 266 | Qwen2.5 7B Instruct · 7.6B | 35.5% | ~5.0 GB |
| 267 | Claude 2.0 · proprietary | 34.7% | API only |
| 268 | Mixtral 8x22B Instruct v0.1 · 140.6B | 34.1% | ~85 GB |
| 269 | Gemini 1.0 Pro 001 · proprietary | 34.0% | API only |
| 270 | Eurus 2 7B PRIME · 7.6B | 33.9% | ~5.0 GB |
| 271 | DeepSeek R1 Distill Qwen 1.5B · 1.8B | 33.6% | ~1.4 GB |
| 272 | Claude 2.1 · proprietary | 33.0% | API only |
| 273 | Gemini 1.5 Flash 8B 001 · proprietary | 33.0% | API only |
| 274 | Dbrx Instruct · proprietary | 32.9% | API only |
| 275 | Yi 1.5 34B Chat · 34.4B | 32.0% | ~21 GB |
| 276 | Qwen1.5 32B Chat · 32.5B | 30.7% | ~20 GB |
| 277 | GPT 4 (Jun 13) · proprietary | 30.6% | API only |
| 278 | Mixtral 8x7B Instruct v0.1 · 46.7B | 30.6% | ~29 GB |
| 279 | Mistral Nemo Instruct 2407 · 12.2B | 29.9% | ~8.1 GB |
| 280 | Qwen1.5 72B Chat · 72.3B | 28.8% | ~49 GB |
| 281 | Granite 4.0 Micro · 3.4B | 28.3% | ~2.5 GB |
| 282 | GPT 3.5 Turbo (Nov 06) · proprietary | 28.0% | API only |
| 283 | Phi 3 Medium 128k Instruct · 14.0B | 27.6% | ~9.1 GB |
| 284 | Gemma 2 9B IT · 9.2B | 27.5% | ~6.1 GB |
| 285 | GPT 3.5 Turbo (Jan 25) · proprietary | 27.2% | API only |
| 286 | Ministral 8B 2410 · proprietary | 27.2% | API only |
| 287 | Llama 3.1 8B Instruct · 8.0B | 27.0% | ~5.3 GB |
| 288 | Llama 2 70B Chat HF · 69.0B | 26.3% | ~152 GB |
| 289 | Meta Llama 3 8B Instruct · 8.0B | 26.1% | ~5.3 GB |
| 290 | Ministral 3B 2410 · proprietary | 25.3% | API only |
| 291 | Deepseek Llm 67B Chat · 67B | 24.6% | ~41 GB |
| 292 | Granite 4.0 1B · 1.6B | 24.0% | ~1.4 GB |
| 293 | Llama 3.2 1B Instruct · 1.2B | 23.9% | ~0.8 GB |
| 294 | Gemma 3 4B IT · 4.3B | 23.2% | ~2.8 GB |
| 295 | Gemma 3 1B IT · 1000M | 20.0% | ~0.7 GB |
| 296 | Mistral 7B Instruct v0.3 · 7.2B | 15.2% | ~4.9 GB |
| 297 | Yi 34B Chat · 34.4B | 14.7% | ~21 GB |
| 298 | Open Mistral 7B · proprietary | 13.2% | API only |
| 299 | Granite 4.0 350M · 352M | 11.2% | ~0.6 GB |
| 300 | DeepSeek R1 0528 Qwen3 8B · 8.2B | 9.3% | ~5.5 GB |
Score vs model size
Which models give the most quality for their size — the ones worth running locally.
- Granite 4.0 350M, 352M, score 11.2% — on the efficiency frontier (best score at its size or smaller).
- Gemma 3 1B IT, 1000M, score 20.0% — on the efficiency frontier (best score at its size or smaller).
- Llama 3.2 1B Instruct, 1B, score 23.9% — on the efficiency frontier (best score at its size or smaller).
- Granite 4.0 1B, 2B, score 24.0% — on the efficiency frontier (best score at its size or smaller).
- DeepSeek R1 Distill Qwen 1.5B, 2B, score 33.6% — on the efficiency frontier (best score at its size or smaller).
- Qwen3 1.7B, 2B, score 38.0% — on the efficiency frontier (best score at its size or smaller).
- Qwen3 4B, 4B, score 52.3% — on the efficiency frontier (best score at its size or smaller).
- Qwen3 8B, 8B, score 56.8% — on the efficiency frontier (best score at its size or smaller).
- Qwen3.5 9B, 10B, score 79.0% — on the efficiency frontier (best score at its size or smaller).
- Qwen3.6 27B, 28B, score 85.9% — on the efficiency frontier (best score at its size or smaller).
- Inkling Small, 266B, score 88.5% — on the efficiency frontier (best score at its size or smaller).
- DeepSeek V4 Flash 0731, 304B, score 91.0% — on the efficiency frontier (best score at its size or smaller).
- GLM 5.2, 753B, score 91.9% — on the efficiency frontier (best score at its size or smaller).
- Kimi K3, 2.8T, score 93.1% — on the efficiency frontier (best score at its size or smaller).
GPQA Diamond: frequently asked questions
- What is the best open LLM on GPQA Diamond?
- Kimi K3 is the top open model on GPQA Diamond, scoring 93.1%. Among all models tested — including proprietary ones — it ranks #21. The top model overall is Gemini 4 Argon (high) (Google DeepMind) at 96.4%.
- What's the best GPQA Diamond model you can run on a 24 GB GPU?
- Qwen3.6 27B is the highest-scoring open model that fits in 24 GB at 4-bit quantization (~17 GB), scoring 85.9% on GPQA Diamond.
- What's the best GPQA Diamond 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 79.0% on GPQA Diamond.
- Can open models match proprietary models on GPQA Diamond?
- Not quite on GPQA Diamond: the strongest proprietary model (Gemini 4 Argon (high)) scores 96.4%, ahead of the best open model (Kimi K3) at 93.1% — 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.