Coding

AA-SciCode leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on AA-SciCode is 66.9% by Claude Opus 5.5. 112 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 5.5	66.9%	measured
2	Claude Fable 5.1	63.1%	measured
3	Gemini 4 Argon	61.8%	measured
4	Claude Fable 5	61.0%	measured
5	Claude Sonnet 5.5	61.0%	measured
6	Claude Mythos 5	61.0%	estimated ± 3.8 pp, high confidence
7	MiMo-V2.6-Pro	60.9%	measured
8	Kimi K3	59.5%	measured
9	GLM-5.3	59.0%	measured
10	Step 5 Preview	58.9%	measured
11	Sakana Fugu	58.8%	estimated ± 3.1 pp, medium confidence
12	Muse Spark 1.1	58.8%	measured
13	Muse Spark 1.3	58.8%	measured
14	Gemini 3.1 Pro	58.7%	measured
15	SWE-2	58.5%	estimated ± 3.8 pp, medium confidence
16	GPT-6 Sol	57.6%	measured
17	Composer 2.5	57.6%	estimated ± 1.7 pp, medium confidence
18	Sakana Fugu-Ultra	57.5%	estimated ± 3.1 pp, high confidence
19	Grok 4.7	57.4%	measured
20	Muse Spark 1.2	57.4%	measured
21	Gemini 3.7 Flash	57.2%	measured
22	GPT-5.6 Sol	57.1%	measured
23	Composer 2	56.6%	estimated ± 8.6 pp, low confidence
24	Gemini 3.8 Flash	56.6%	measured
25	GPT-6 Astra	56.5%	measured
26	Grok 4.6	56.5%	measured
27	Claude Opus 5	56.4%	measured
28	Composer 2 Fast	56.1%	estimated ± 8.6 pp, low confidence
29	SWE-1.7	56.0%	estimated ± 3.8 pp, medium confidence
30	Claude Opus 4.7	55.8%	estimated ± 3.5 pp, high confidence
31	GPT-5.5	55.8%	measured
32	Claude Opus 4.7 (Adaptive)	55.4%	estimated ± 3.2 pp, high confidence
33	Claude Haiku 5.5	55.0%	measured
34	GPT-5.6 Terra	55.0%	measured
35	Grok 4.5	55.0%	measured
36	GPT-6 Luna	54.6%	measured
37	Claude Opus 4.8	54.4%	measured
38	Claude Sonnet 5	54.3%	measured
39	GPT-5.4	54.3%	estimated ± 3.2 pp, high confidence
40	GPT-6.1 Sol	54.2%	measured
41	Mistral Large 4	54.2%	measured
42	Ling 3.1 Flash	54.1%	measured
43	Gemini 3.5 Flash	53.9%	measured
44	Ember-1	53.6%	estimated ± 3.9 pp, high confidence
45	GPT-5.6 Luna	53.6%	measured
46	Claude Opus 4.6 (Adaptive)	53.5%	estimated ± 3.8 pp, high confidence
47	Gemini 3.6 Flash	53.4%	measured
48	Ornith-1.5-397B	52.8%	estimated ± 3.0 pp, high confidence
49	Hy4 preview	52.7%	estimated ± 3.0 pp, high confidence
50	Qwen3.8 Max	52.2%	estimated ± 3.0 pp, high confidence
51	Claude Sonnet 4.6	52.2%	estimated ± 1.7 pp, medium confidence
52	GPT-5.4 mini	52.1%	measured
53	Qwen3.8 Max Preview	52.1%	measured
54	DeepSeek V4.1 Flash	51.9%	measured
55	GLM-5.3-Flash	51.6%	measured
56	Kimi K2.6	51.5%	measured
57	MiMo-V2.6-Flash	51.3%	measured
58	GPT-5.2-Codex	51.2%	estimated ± 3.8 pp, high confidence
59	GLM-5.2	51.2%	measured
60	DeepSeek V4 Pro 0813	51.0%	measured
61	dots3-note Preview	50.8%	estimated ± 3.0 pp, high confidence
62	MiMo-V2.5-Pro	50.6%	measured
63	Qwen3.8-Flash-Next	50.6%	measured
64	Qwen3.8-Omni-Flash	50.5%	estimated ± 3.0 pp, high confidence
65	DeepSeek V4 Flash 0731	50.3%	measured
66	Ornith-1.0-397B	50.2%	estimated ± 3.0 pp, high confidence
67	Atria Dawn Preview	50.1%	estimated ± 3.8 pp, high confidence
68	MiniMax M2.7	50.1%	measured
69	Laguna S 2.1	50.0%	estimated ± 3.8 pp, high confidence
70	Seed 2.1 Pro	49.8%	estimated ± 3.0 pp, high confidence
71	Inkling-Small	49.7%	measured
72	GLM-4.5	49.6%	estimated ± 4.6 pp, low confidence
73	Qwen3.7 Max	49.5%	measured
74	Ornith-1.5-35B-A3B	49.5%	estimated ± 3.0 pp, high confidence
75	Beam	49.2%	estimated ± 3.1 pp, high confidence
76	GLM-5 (Reasoning)	49.1%	estimated ± 3.8 pp, high confidence
77	Claude Sonnet 4.5 Thinking	49.0%	estimated ± 3.8 pp, high confidence
78	GPT-5.1-Codex-Max	49.0%	estimated ± 3.8 pp, high confidence
79	Claude Opus 4.5 Thinking	48.7%	estimated ± 3.8 pp, high confidence
80	Muse Spark	48.7%	estimated ± 3.2 pp, high confidence
81	GPT-5.3 Codex	48.7%	estimated ± 3.8 pp, high confidence
82	Gemini 3 Flash	48.7%	estimated ± 3.8 pp, high confidence
83	Solar Open 2	48.6%	estimated ± 3.7 pp, medium confidence
84	Hy3	48.6%	measured
85	Hy3 Preview	48.6%	measured
86	MiMo-V2.5	48.3%	estimated ± 3.8 pp, high confidence
87	Grok 4.3	48.3%	measured
88	Seed 2.1 Turbo	48.3%	estimated ± 3.0 pp, high confidence
89	Kimi K2.5	48.3%	estimated ± 3.1 pp, high confidence
90	Quasar 438B	48.1%	measured
91	GPT-5.2	48.1%	estimated ± 3.8 pp, high confidence
92	Claude Opus 4.5	48.0%	estimated ± 3.0 pp, high confidence
93	Qwen3.5 Plus	48.0%	estimated ± 3.8 pp, high confidence
94	MiniMax M2.5	47.9%	estimated ± 3.8 pp, high confidence
95	Qwen 3.6 Max (preview)	47.9%	estimated ± 3.0 pp, high confidence
96	Gemini 3 Pro	47.8%	estimated ± 3.8 pp, high confidence
97	Kimi K2.7 Code	47.8%	measured
98	GPT-5 mini	47.8%	estimated ± 3.8 pp, high confidence
99	GLM-5	47.8%	estimated ± 3.8 pp, high confidence
100	GPT-5.1-Codex	47.6%	estimated ± 3.8 pp, high confidence
101	Claude Haiku 4.5 Thinking	47.4%	estimated ± 3.8 pp, high confidence
102	MiMo-V2-Pro	47.3%	estimated ± 3.9 pp, high confidence
103	Ternary Bonsai 2 27B	47.2%	estimated ± 3.7 pp, high confidence
104	GPT-5.4 nano	47.2%	measured
105	MiniMax M3	47.1%	measured
106	Inkling	47.0%	measured
107	Claude Sonnet 4.5	46.9%	estimated ± 3.9 pp, high confidence
108	Claude Opus 4.6	46.9%	estimated ± 3.8 pp, high confidence
109	Qwen3.6 Plus	46.9%	estimated ± 3.2 pp, high confidence
110	Qwen3.8-27B	46.6%	measured
111	DeepSeek V3.2 (Thinking)	46.5%	estimated ± 3.8 pp, high confidence
112	Gemini 2.5 Pro	46.3%	measured
113	Qwen3 Max	46.2%	estimated ± 3.8 pp, high confidence
114	GLM-4.6	46.2%	estimated ± 3.8 pp, high confidence
115	Qwen3.7 Plus	46.1%	measured
116	Grok 4.20	46.1%	estimated ± 3.8 pp, high confidence
117	Grok 4	46.0%	estimated ± 8.6 pp, low confidence
118	Grok 4.1 Fast (Reasoning)	45.9%	estimated ± 3.8 pp, high confidence
119	BTL-3	45.9%	estimated ± 3.7 pp, high confidence
120	MiMo-V2-Omni	45.8%	estimated ± 3.9 pp, high confidence
121	Gemini 3.1 Flash-Lite	45.7%	estimated ± 3.8 pp, medium confidence
122	Grok 4 Fast (Reasoning)	45.7%	estimated ± 3.8 pp, medium confidence
123	Claude 4.1 Opus	45.7%	estimated ± 3.9 pp, high confidence
124	MAI-Thinking-1	45.6%	estimated ± 3.7 pp, high confidence
125	Apodex 1.1	45.5%	measured
126	Apodex 1.1 Mini	45.5%	measured
127	Gemma 4 31B	45.5%	measured
128	DeepSeek V3.2	45.4%	estimated ± 8.6 pp, low confidence
129	Qwen3.5 Flash	45.2%	estimated ± 4.3 pp, medium confidence
130	Claude Haiku 4.5	45.1%	estimated ± 3.9 pp, high confidence
131	Muse Glimmer 30B	44.9%	measured
132	Claude 4 Sonnet	44.8%	estimated ± 3.9 pp, high confidence
133	GLM-5.1	44.8%	measured
134	MiMo-V2-Flash	44.8%	estimated ± 3.2 pp, high confidence
135	MAI-Code-1.1-Flash	44.8%	estimated ± 3.9 pp, high confidence
136	Laguna M.1	44.7%	estimated ± 3.8 pp, high confidence
137	Qwen3.5-27B	44.7%	estimated ± 3.9 pp, high confidence
138	Solar Pro 4	44.6%	measured
139	GPT-5.1	44.6%	estimated ± 3.2 pp, high confidence
140	Ling 3.0 Flash VL	44.2%	measured
141	Step 3.7 Flash	43.9%	measured
142	Grok Code Fast 1	43.9%	estimated ± 3.9 pp, high confidence
143	Laguna XS 2.1	43.9%	estimated ± 3.8 pp, high confidence
144	Kimi K2.5 (Reasoning)	43.4%	estimated ± 3.2 pp, high confidence
145	Laguna XS.2	43.2%	estimated ± 3.8 pp, high confidence
146	Qwen3.5-35B-A3B	43.1%	estimated ± 3.9 pp, high confidence
147	Qwen3.6-27B	42.8%	measured
148	GLM-4.7	42.7%	estimated ± 3.2 pp, high confidence
149	Qwen3.5 397B	42.3%	estimated ± 3.7 pp, high confidence
150	K-EXAONE 2.0	42.0%	measured
151	Ling 3.0 Flash	42.0%	measured
152	Ling 3.0 Flash FP8	42.0%	measured
153	Ornith-1.0-35B	41.8%	estimated ± 3.0 pp, high confidence
154	Gemini 3.5 Flash-Lite	41.3%	measured
155	A.X K2	41.0%	measured
156	Trinity-Large-Preview	40.6%	measured
157	Trinity-Large-Thinking	40.6%	measured
158	Nemotron 3 Ultra	40.3%	measured
159	o1	40.2%	estimated ± 3.2 pp, high confidence
160	LongCat-Flash-Lite-Sparse	40.2%	estimated ± 3.8 pp, high confidence
161	Mistral Medium 3.5 128B	40.2%	measured
162	Gemma 4 26B A4B	40.0%	measured
163	Qwen3.5-122B-A10B	39.7%	measured
164	Ornith-1.5-9B	39.7%	estimated ± 3.0 pp, high confidence
165	GPT-5 (high)	39.4%	estimated ± 3.2 pp, high confidence
166	DeepSeek V3 0324	39.0%	measured
167	GPT-OSS 20B	38.9%	measured
168	Mistral Small 4	38.8%	measured
169	Mistral Small 4 (Reasoning)	38.8%	measured
170	North Mini Code	38.8%	measured
171	Command A+	38.5%	measured
172	Mercury 2.5	38.5%	measured
173	Granite 4.2 30B	37.8%	measured
174	o1-preview	37.7%	estimated ± 3.2 pp, high confidence
175	K-Exaone	36.8%	estimated ± 3.2 pp, high confidence
176	Mistral Large 3	36.6%	measured
177	Qwen3.6-35B-A3B	36.6%	measured
178	Gemma 4 12B	36.3%	estimated ± 3.2 pp, high confidence
179	Nemotron 3 Super 100B	36.2%	measured
180	DeepSeek V3	35.8%	measured
181	GPT-4.1	35.6%	estimated ± 3.9 pp, high confidence
182	LLaDA2.2-flash	34.7%	estimated ± 3.8 pp, high confidence
183	Ornith-1.0-9B	34.6%	estimated ± 3.0 pp, medium confidence
184	GPT-OSS 120B	34.0%	measured
185	Gemini 1.5 Pro	33.1%	estimated ± 3.2 pp, high confidence
186	Nemotron 3 Nano Omni 30B A3B	32.9%	estimated ± 3.1 pp, high confidence
187	o3-mini	32.8%	estimated ± 3.9 pp, high confidence
188	Claude 3.5 Sonnet	32.6%	estimated ± 3.9 pp, high confidence
189	Nemotron 3.5 Lightning 30B A3B NVFP4	32.1%	measured
190	GPT-4 Turbo	32.1%	estimated ± 3.2 pp, high confidence
191	Llama 4 Maverick	31.7%	measured
192	GPT-4.1 mini	31.5%	estimated ± 3.2 pp, high confidence
193	Granite 4.2 8B	31.5%	measured
194	Claude 3 Opus	31.2%	estimated ± 3.2 pp, high confidence
195	Agents-A1-4B	30.6%	estimated ± 3.1 pp, high confidence
196	Nemotron 3 Nano 30B	30.6%	measured
197	Ling 2.6 Flash	28.2%	estimated ± 3.1 pp, high confidence
198	Ultravox v0.6 Llama 3.3 70B	27.8%	estimated ± 3.2 pp, high confidence
199	GPT-4o mini	27.6%	estimated ± 3.2 pp, high confidence
200	GPT-4.1 nano	27.5%	estimated ± 3.2 pp, high confidence
201	Mellum2-12B-A2.5B-Thinking	27.3%	estimated ± 3.7 pp, high confidence
202	MiniCPM5-2B	26.3%	measured
203	Gemma 4 E2B	25.7%	estimated ± 3.2 pp, medium confidence
204	Solar Pro 3	25.5%	measured
205	Granite 4.2 3B	25.3%	measured
206	Gemma 4 E4B	24.4%	measured
207	Ling 3.0 Tiny	24.2%	measured
208	Gemma 3 27B	23.3%	measured
209	BTL-4	22.5%	estimated ± 3.7 pp, high confidence
210	ZAYA1-8B	22.1%	estimated ± 3.7 pp, high confidence
211	ZAYA1-74B-Preview	21.9%	estimated ± 3.7 pp, high confidence
212	Celeris-1	21.6%	measured
213	Llama 4 Scout	21.3%	measured
214	LFM2.5-2.6B	14.4%	measured
215	Mellum2-12B-A2.5B-Instruct	1.1%	estimated ± 3.7 pp, medium confidence
216	MiniCPM5-1B	0.6%	estimated ± 3.7 pp, medium confidence
217	LLaDA2.2-mini	0.2%	estimated ± 3.7 pp, medium confidence

AA-SciCode leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General