benchgap
Coding

AA-SciCode leaderboard

As of 2026-11-21, the highest measured score on AA-SciCode is 67.7% by Claude Fable 5.2. 169 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Gemini 4 Neon71.9%estimated ± 2.2 pp, medium confidence
2Claude Fable 5.267.7%measured
3Gemini 4 Carbon67.7%estimated ± 2.2 pp, medium confidence
4Gemini 4 Hydrogen67.6%estimated ± 2.2 pp, medium confidence
5Claude Opus 5.566.9%measured
6Claude Fable 5.163.1%measured
7Muse Wonder Full62.6%measured
8Gemini 4 Argon61.8%measured
9GPT-6.2 Astra61.7%estimated ± 3.5 pp, medium confidence
10GPT-6.2 Sol61.4%estimated ± 3.5 pp, medium confidence
11Claude Fable 561.0%measured
12Claude Sonnet 5.561.0%measured
13MiMo-V2.6-Pro60.9%measured
14Kimi K3.1 Pro60.8%measured
15Claude Mythos 560.8%estimated ± 3.6 pp, high confidence
16Muse Wonder60.3%measured
17Hy460.1%estimated ± 3.0 pp, medium confidence
18Kimi K359.5%measured
19GLM-5.359.0%measured
20Qwen4 Max59.0%estimated ± 3.9 pp, medium confidence
21Step 5 Preview58.9%measured
22Muse Spark 1.158.8%measured
23Muse Spark 1.358.8%measured
24Sakana Fugu58.7%estimated ± 3.0 pp, medium confidence
25Gemini 3.1 Pro58.7%measured
26Pareto 26.958.6%estimated ± 3.9 pp, medium confidence
27MAI Code 258.5%estimated ± 3.9 pp, medium confidence
28Muse Spark Open58.4%measured
29Particle-158.3%estimated ± 3.9 pp, medium confidence
30Pareto 26.10 Preview57.7%estimated ± 3.9 pp, medium confidence
31GPT-6 Sol57.6%measured
32Composer 2.557.6%estimated ± 1.7 pp, medium confidence
33Sakana Fugu-Ultra57.4%estimated ± 3.0 pp, high confidence
34Grok 4.757.4%measured
35Muse Spark 1.257.4%measured
36Gemini 3.7 Flash57.2%measured
37GPT-5.6 Sol57.1%measured
38SWE-256.7%estimated ± 3.5 pp, medium confidence
39SWE-2 Flash56.7%estimated ± 3.5 pp, medium confidence
40Mistral Large 456.6%estimated ± 3.5 pp, medium confidence
41Composer 256.6%estimated ± 8.6 pp, low confidence
42SWE-1.756.6%estimated ± 3.5 pp, medium confidence
43Inkling 2 Mini56.6%estimated ± 3.5 pp, medium confidence
44Gemini 3.8 Flash56.6%measured
45GPT-6 Astra56.5%measured
46Grok 4.656.5%measured
47Claude Opus 556.4%measured
48Hy4 flash56.4%estimated ± 3.9 pp, medium confidence
49Composer 2 Fast56.1%estimated ± 8.6 pp, low confidence
50GPT-5.555.8%measured
51Ultramarine 1 Flash55.5%estimated ± 3.9 pp, medium confidence
52Claude Opus 4.755.5%estimated ± 3.4 pp, high confidence
53GPT-6.1 Terra55.1%measured
54MAI Code 2 Flash55.1%estimated ± 3.9 pp, medium confidence
55Claude Haiku 5.555.0%measured
56GPT-5.6 Terra55.0%measured
57Grok 4.555.0%measured
58GLM-5.3-Flash54.8%estimated ± 2.5 pp, high confidence
59GPT-6 Luna54.6%measured
60Claude Opus 4.854.4%measured
61ERNIE 5.5 Flash Preview54.4%estimated ± 3.1 pp, high confidence
62GPT-5.454.3%estimated ± 3.4 pp, high confidence
63Claude Sonnet 554.3%measured
64GPT-6.1 Sol54.2%measured
65Ling 3.1 Flash54.1%measured
66Claude Opus 4.6 (Adaptive)54.0%estimated ± 3.8 pp, high confidence
67Gemini 3.5 Flash53.9%measured
68Ember-153.8%estimated ± 3.8 pp, high confidence
69GPT-5.6 Luna53.6%measured
70Gemini 3.6 Flash53.4%measured
71MiniMax M3.1 Flash53.4%measured
72LongCat-2.553.3%estimated ± 3.1 pp, high confidence
73Ornith-1.5-397B53.0%estimated ± 3.1 pp, high confidence
74Inkling 2 Small52.3%measured
75Claude Sonnet 4.652.2%estimated ± 1.7 pp, medium confidence
76GPT-5.4 mini52.1%measured
77Qwen3.8 Max Preview52.1%measured
78Qwen3.8 Max52.0%estimated ± 3.1 pp, high confidence
79DeepSeek V4.1 Flash 101751.9%measured
80o351.8%estimated ± 5.1 pp, low confidence
81Kimi K2.651.5%measured
82GPT-5.2-Codex51.4%estimated ± 3.8 pp, high confidence
83MiMo-V2.6-Flash51.3%measured
84GLM-5.251.2%measured
85o4-mini51.0%estimated ± 5.1 pp, low confidence
86DeepSeek V4 Pro51.0%measured
87MiMo-V2.5-Pro50.6%measured
88Qwen3.8-Flash-Next50.6%measured
89Mercury 350.4%estimated ± 3.9 pp, medium confidence
90DeepSeek V4 Flash 073150.3%measured
91MiniMax M2.750.1%measured
92Atria Dawn Preview50.1%estimated ± 3.6 pp, high confidence
93dots3-note Preview50.1%estimated ± 3.1 pp, high confidence
94Laguna S 2.150.0%estimated ± 3.6 pp, high confidence
95Qwen3.8-Omni-Flash49.7%estimated ± 3.1 pp, high confidence
96Inkling-Small49.7%measured
97Qwen3.7 Max49.5%measured
98Ornith-1.0-397B49.5%estimated ± 3.1 pp, high confidence
99Seed 2.1 Pro49.1%estimated ± 3.1 pp, high confidence
100Claude Sonnet 4.5 Thinking48.9%estimated ± 3.8 pp, high confidence
101GPT-5.1-Codex-Max48.9%estimated ± 3.8 pp, high confidence
102Ornith-1.5-35B-A3B48.8%estimated ± 3.1 pp, high confidence
103Muse Spark48.7%estimated ± 3.4 pp, high confidence
104GPT-5.3 Codex48.6%estimated ± 3.6 pp, high confidence
105Claude Opus 4.5 Thinking48.6%estimated ± 3.8 pp, high confidence
106Hy348.6%measured
107Hy3 Preview48.6%measured
108Gemini 3 Flash48.5%estimated ± 3.8 pp, high confidence
109Beam 1.0 501B-A23B48.5%measured
110Grok 4.348.3%measured
111MiMo-V2.548.3%estimated ± 3.6 pp, high confidence
112Kimi K2.548.2%estimated ± 3.0 pp, high confidence
113Quasar 438B48.1%measured
114GPT-5.248.0%estimated ± 3.6 pp, high confidence
115Solar Pro 448.0%measured
116Beam 1.0 501B-A23B FP847.8%estimated ± 3.0 pp, high confidence
117Qwen3.5 Plus47.8%estimated ± 3.8 pp, high confidence
118Kimi K2.7 Code47.8%measured
119Nemotron 3.5 Super VL 120B A12B47.8%measured
120Nemotron 3.5 Super VL 120B A12B NVFP447.8%measured
121Seed 2.1 Turbo47.8%estimated ± 3.1 pp, high confidence
122BTL-447.7%estimated ± 3.8 pp, high confidence
123MiniMax M2.547.7%estimated ± 3.8 pp, high confidence
124Gemini 3 Pro47.6%estimated ± 3.8 pp, high confidence
125Claude Opus 4.547.6%estimated ± 3.1 pp, high confidence
126GPT-5 mini47.6%estimated ± 3.8 pp, high confidence
127MiMo-V2-Pro47.5%estimated ± 3.8 pp, high confidence
128GPT-5.1-Codex47.4%estimated ± 3.8 pp, high confidence
129GPT-5.4 nano47.2%measured
130Claude Sonnet 4.547.1%estimated ± 3.8 pp, high confidence
131MiniMax M347.1%measured
132Claude Haiku 4.5 Thinking47.1%estimated ± 3.8 pp, high confidence
133Inkling47.0%measured
134Claude Opus 4.646.9%estimated ± 3.6 pp, high confidence
135Qwen 3.6 Max (preview)46.6%estimated ± 3.0 pp, high confidence
136Qwen3.8-27B46.6%measured
137MAI-Thinking-146.6%estimated ± 3.6 pp, high confidence
138Beam 1.0 501B-A23B NVFP446.5%estimated ± 3.0 pp, high confidence
139BTL-346.4%estimated ± 4.7 pp, high confidence
140Gemini 2.5 Pro46.3%measured
141Qwen3.7 Plus46.1%measured
142Grok 4.2046.0%estimated ± 3.6 pp, high confidence
143MiMo-V2-Omni46.0%estimated ± 3.8 pp, high confidence
144Claude 4.1 Opus45.9%estimated ± 3.8 pp, high confidence
145Step 3.5 Flash45.8%estimated ± 3.8 pp, high confidence
146Qwen3 Max45.8%estimated ± 3.8 pp, high confidence
147GLM-4.645.7%estimated ± 3.8 pp, high confidence
148Flint45.7%estimated ± 3.8 pp, high confidence
149Apodex 1.145.5%measured
150Apodex 1.1 Mini45.5%measured
151Gemma 4 31B45.5%measured
152Grok 4.1 Fast (Reasoning)45.4%estimated ± 3.8 pp, high confidence
153DeepSeek V3.245.4%estimated ± 8.6 pp, low confidence
154Claude Haiku 4.545.3%estimated ± 3.8 pp, high confidence
155Gemini 3.1 Flash-Lite45.2%estimated ± 3.8 pp, medium confidence
156Grok 4 Fast (Reasoning)45.2%estimated ± 3.8 pp, medium confidence
157Qwen3.5 Flash45.2%estimated ± 4.2 pp, high confidence
158Qwen3.6 Plus45.2%estimated ± 3.1 pp, high confidence
159Jais-3-MoVA-36B-A4B45.1%estimated ± 3.8 pp, high confidence
160Claude 4 Sonnet45.0%estimated ± 3.8 pp, high confidence
161MAI-Code-1.1-Flash45.0%estimated ± 3.8 pp, high confidence
162Claude 4 Opus44.9%estimated ± 3.8 pp, high confidence
163Muse Glimmer 30B44.9%measured
164Jais-3-32B44.9%estimated ± 3.8 pp, high confidence
165Qwen3.5-27B44.9%estimated ± 3.8 pp, high confidence
166GLM-5.144.8%measured
167Qwen3.5 397B44.8%measured
168MiMo-V2-Flash44.8%estimated ± 3.4 pp, high confidence
169Laguna M.144.7%estimated ± 3.6 pp, high confidence
170GPT-5.144.6%estimated ± 3.4 pp, high confidence
171Ling 3.0 Flash VL44.2%measured
172GLM-544.2%estimated ± 3.1 pp, high confidence
173Sage-A1.544.1%estimated ± 3.0 pp, high confidence
174Grok Code Fast 144.1%estimated ± 3.8 pp, high confidence
175Step 3.7 Flash43.9%measured
176Solar Open 243.9%estimated ± 3.8 pp, high confidence
177Laguna XS 2.143.9%estimated ± 3.6 pp, high confidence
178GLM-4.543.8%estimated ± 5.1 pp, low confidence
179Jais-3-375B-A23B43.8%estimated ± 3.8 pp, high confidence
180Jais-3-7B43.8%estimated ± 3.8 pp, high confidence
181Qwen3-Coder 480B-A35B43.5%estimated ± 3.8 pp, high confidence
182Ornith-1.0-35B43.5%estimated ± 3.1 pp, high confidence
183Qwen3.5-35B-A3B43.3%estimated ± 3.8 pp, high confidence
184Laguna XS.243.2%estimated ± 3.6 pp, high confidence
185Qwen3.6-27B42.8%measured
186GLM-4.742.7%estimated ± 3.4 pp, high confidence
187Grok 442.7%estimated ± 4.2 pp, high confidence
188Ornith-1.5-9B42.2%estimated ± 3.1 pp, high confidence
189K-EXAONE 2.042.0%measured
190Ling 3.0 Flash42.0%measured
191DeepSeek V3.141.7%estimated ± 3.8 pp, high confidence
192Kimi K241.6%estimated ± 3.8 pp, high confidence
193Gemini 3.5 Flash-Lite41.3%measured
194A.X K241.0%measured
195Trinity-Large-Preview40.6%measured
196Trinity-Large-Thinking40.6%measured
197Ling 3.0 Flash FP840.5%estimated ± 3.0 pp, high confidence
198Nemotron 3 Ultra40.3%measured
199LongCat-Flash-Lite-Sparse40.3%estimated ± 3.6 pp, high confidence
200Mistral Medium 3.5 128B40.2%measured
201Claude 3.7 Sonnet Thinking40.2%estimated ± 5.1 pp, low confidence
202o140.2%estimated ± 3.4 pp, high confidence
203Gemma 4 26B A4B40.0%measured
204Claude 3.7 Sonnet39.8%estimated ± 3.8 pp, high confidence
205Nemotron 3.5 Bolt 74B-A9B39.7%measured
206Qwen3.5-122B-A10B39.7%measured
207Ripple-65B Preview39.6%estimated ± 3.0 pp, high confidence
208GPT-539.3%estimated ± 3.4 pp, high confidence
209Ternary Bonsai 2 27B39.1%estimated ± 3.8 pp, high confidence
210DeepSeek V3 032439.0%measured
211Ornith-1.0-9B38.9%estimated ± 3.1 pp, medium confidence
212Nemotron 3 Super 100B38.9%estimated ± 3.8 pp, high confidence
213GPT-OSS 20B38.9%measured
214Mistral Small 438.8%measured
215North Mini Code38.8%measured
216Command A+38.5%measured
217Mercury 2.538.5%measured
218GLM-4.7-Flash38.2%estimated ± 3.8 pp, high confidence
219Granite 4.2 30B37.8%measured
220o1-preview37.6%estimated ± 3.4 pp, high confidence
221DeepSeek-R1-052837.4%estimated ± 3.8 pp, high confidence
222Spark-X2.6-4B37.2%estimated ± 3.0 pp, high confidence
223K-Exaone36.8%estimated ± 3.4 pp, high confidence
224Mistral Large 336.6%measured
225Qwen3.6-35B-A3B36.6%measured
226MiniMax M1 80k36.6%estimated ± 3.8 pp, high confidence
227Gemma 4 12B36.2%estimated ± 3.4 pp, high confidence
228GPT-4.135.8%estimated ± 3.8 pp, high confidence
229DeepSeek V335.8%measured
230ZAYA1-74B-Preview35.0%estimated ± 3.8 pp, high confidence
231LLaDA2.2-flash34.8%estimated ± 3.6 pp, high confidence
232GPT-OSS 120B34.0%measured
233Mellum2.1-12B-A2.5B-Thinking33.7%estimated ± 3.6 pp, high confidence
234LFM2.5-24B-A2B33.4%estimated ± 3.0 pp, high confidence
235Gemini 1.5 Pro32.9%estimated ± 3.4 pp, high confidence
236o3-mini32.9%estimated ± 3.8 pp, high confidence
237DeepSeek-R132.8%estimated ± 3.8 pp, high confidence
238Nemotron 3 Nano Omni 30B A3B32.8%estimated ± 3.0 pp, high confidence
239Claude 3.5 Sonnet32.7%estimated ± 3.8 pp, high confidence
240Nemotron 3.5 Lightning 30B A3B NVFP432.1%measured
241GPT-4 Turbo32.0%estimated ± 3.4 pp, high confidence
242Llama 4 Maverick31.7%measured
243Granite 4.2 8B31.5%measured
244Devstral Small31.5%estimated ± 3.8 pp, high confidence
245GPT-4.1 mini31.4%estimated ± 3.4 pp, high confidence
246Claude 3 Opus31.1%estimated ± 3.4 pp, high confidence
247GPT-4o30.6%estimated ± 5.1 pp, low confidence
248Agents-A1-4B30.5%estimated ± 3.0 pp, high confidence
249Sarvam 105B30.5%estimated ± 3.8 pp, high confidence
250Mellum2-12B-A2.5B-Thinking30.1%estimated ± 4.7 pp, high confidence
251LFM3-1.7B29.5%estimated ± 3.0 pp, high confidence
252Nemotron 3 Nano 30B28.8%estimated ± 3.4 pp, high confidence
253Ling 2.6 Flash28.1%estimated ± 3.0 pp, high confidence
254Claude 3.5 Haiku27.9%estimated ± 3.8 pp, medium confidence
255Mistral Small 3.127.8%measured
256Ultravox v0.6 Llama 3.3 70B27.7%estimated ± 3.4 pp, high confidence
257GPT-4o mini27.4%estimated ± 3.4 pp, high confidence
258GPT-4.1 nano27.3%estimated ± 3.4 pp, high confidence
259ZAYA1-8B27.0%estimated ± 4.7 pp, high confidence
260MiniCPM5-2B26.3%measured
261Gemma 4 E2B25.5%estimated ± 3.4 pp, medium confidence
262Solar Pro 325.5%measured
263Granite 4.2 3B25.3%measured
264Gemma 4 E4B24.4%measured
265Ling 3.0 Tiny24.2%measured
266Sarvam 30B23.9%estimated ± 3.8 pp, medium confidence
267Gemma 3 27B23.3%measured
268Gemini 2.5 Flash-Lite22.4%estimated ± 3.8 pp, medium confidence
269Celeris-121.6%measured
270Llama 4 Scout21.3%measured
271Qwen3 235B 250717.5%estimated ± 4.7 pp, medium confidence
272Magistral Medium16.5%estimated ± 4.7 pp, medium confidence
273Gemini 2.0 Flash15.7%estimated ± 3.8 pp, medium confidence
274Magistral Small14.8%estimated ± 4.7 pp, medium confidence
275LFM2.5-2.6B14.4%measured
276Gemini 2.0 Flash-Lite9.5%estimated ± 3.8 pp, medium confidence
277Mellum2-12B-A2.5B-Instruct9.1%estimated ± 4.7 pp, medium confidence
278Gemini 1.5 Flash7.4%estimated ± 3.8 pp, medium confidence
279MiniCPM5-1B7.2%estimated ± 4.7 pp, medium confidence
280Hy4 preview6.9%estimated ± 3.0 pp, low confidence
281Mistral Medium 35.6%estimated ± 4.7 pp, medium confidence
282LLaDA2.2-mini4.5%estimated ± 4.7 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General