Skip to main content
BenchLM
Advertise

Benchmark profile

Artificial Analysis SciCode (AA-SciCode)

A display-only Artificial Analysis SciCode score.

Data verified

Benchmark score on AA-SciCode — July 23, 2026

BenchLM mirrors the published score view for AA-SciCode. Claude Fable 5 leads the public snapshot at 60.2% , followed by Gemini 3.1 Pro (58.9%) and Kimi K3 (58.7%). BenchLM does not use these results to rank models overall.

1Closed

Claude Fable 5

Anthropic

claude-fable-5

60.2%
Overall 83.68Context 1M+
2Closed

Gemini 3.1 Pro

Google

gemini-3-1-pro

58.9%
Overall 55.3Context 1M
3Closed

Kimi K3

Moonshot AI

kimi-3

58.7%
Overall 80.96Context 1.05M
162 modelsCodingCurrentDisplay onlyUpdated July 23, 2026

Benchmark score table (162 models)

Score
1
Claude Fable 5Anthropic · Closed
60.2%
2
Gemini 3.1 ProGoogle · Closed
58.9%
3
Kimi K3Moonshot AI · Closed
58.7%
4
Muse Spark 1.1Meta · Closed
58.2%
5
GPT-5.4OpenAI · Closed
56.6%
6
GPT-5.6 SolOpenAI · Closed
56.1%
7
GPT-5.5OpenAI · Closed
56.1%
8
Gemini 3 ProGoogle · Closed
56.1%
9
GPT-5.2-CodexOpenAI · Closed
54.6%
10
Claude Opus 4.7 (Adaptive)Anthropic · Closed
54.5%
11
Grok 4.5xAI · Closed
54.1%
12
GPT-5.6 TerraOpenAI · Closed
53.9%
13
Claude Sonnet 5Anthropic · Closed
53.6%
14
Claude Opus 4.8Anthropic · Closed
53.5%
15
Kimi K2.6Moonshot AI · Open weight
53.5%
16
GPT-5.3 CodexOpenAI · Closed
53.2%
17
GPT-5.3-Codex-SparkOpenAI · Closed
53.2%
18
Gemini 3.5 FlashGoogle · Closed
53.1%
19
Gemini 3.6 FlashGoogle · Closed
52.7%
20
GPT-5.6 LunaOpenAI · Closed
52.5%
21
GPT-5.2OpenAI · Closed
52.1%
22
Claude Opus 4.6 (Adaptive)Anthropic · Closed
51.9%
23
Muse SparkMeta · Closed
51.5%
24
GLM-5.2Z.AI · Open weight
50.5%
25
MiMo-V2.5-ProXiaomi · Closed
50.2%
26
Claude Opus 4.7Anthropic · Closed
50.1%
27
DeepSeek V4 Pro (Max)DeepSeek · Open weight
50.0%
28
GPT-5.4 miniOpenAI · Closed
49.9%
29
Gemini 3 FlashGoogle · Closed
49.9%
30
Claude Opus 4.5 ThinkingAnthropic · Closed
49.5%
31
Kimi K2.5Moonshot AI · Open weight
49.0%
32
Kimi K2.5 (Reasoning)Moonshot AI · Closed
49.0%
33
Qwen3.7 MaxAlibaba · Closed
48.8%
34
Hy3 PreviewTencent · Open weight
47.6%
35
Hy3Tencent · Open weight
47.6%
36
Kimi K2.7 CodeMoonshot AI · Open weight
47.5%
37
Grok 4.3xAI · Closed
47.3%
38
Claude Opus 4.5Anthropic · Closed
47.0%
39
MiniMax M2.7MiniMax · Open weight
47.0%
40
Claude Sonnet 4.6Anthropic · Closed
46.9%
41
GPT-5.4 nanoOpenAI · Closed
46.9%
42
Qwen 3.6 Max (preview)Alibaba · Closed
46.9%
43
DeepSeek V4 Pro (High)DeepSeek · Open weight
46.4%
44
GLM-5Z.AI · Open weight
46.2%
45
InklingThinking Machines Lab · Open weight
46.1%
46
Claude Opus 4.6Anthropic · Closed
45.7%
47
Grok 4xAI · Closed
45.7%
48
Qwen3.7 PlusAlibaba · Closed
45.5%
49
MiniMax M3MiniMax · Open weight
45.4%
50
GLM-4.7Z.AI · Open weight
45.1%
51
DeepSeek V4 Flash (Max)DeepSeek · Open weight
44.9%
52
Grok 4.1 Fast (Reasoning)xAI · Closed
44.2%
53
Grok 4 Fast (Reasoning)xAI · Closed
44.2%
54
GLM-5.1Z.AI · Open weight
43.8%
55
GLM-5-TurboZ.AI · Closed
43.6%
56
GLM-5V-TurboZ.AI · Closed
43.5%
57
Gemma 4 31BGoogle · Open weight
43.4%
58
GPT-5.1OpenAI · Closed
43.3%
59
GPT-5 (high)OpenAI · Closed
42.9%
60
Gemini 2.5 ProGoogle · Closed
42.8%
61
MiniMax M2.5MiniMax · Closed
42.6%
62
MiMo-V2-ProXiaomi · Closed
42.5%
63
Qwen3.5 397BAlibaba · Open weight
42.0%
64
Qwen3.5-122B-A10BAlibaba · Open weight
42.0%
65
DeepSeek V4 Flash (High)DeepSeek · Open weight
42.0%
66
Qwen3.5 397B (Reasoning)Alibaba · Open weight
42.0%
67
Gemini 3.1 Flash-LiteGoogle · Closed
41.9%
68
GPT-5 (medium)OpenAI · Closed
41.1%
69
o3OpenAI · Closed
41.0%
70
Gemini 3.5 Flash-LiteGoogle · Closed
40.9%
71
Claude 4.1 Opus ThinkingAnthropic · Closed
40.9%
72
Qwen3.6 PlusAlibaba · Closed
40.7%
73
GPT-4.1 miniOpenAI · Closed
40.4%
74
DeepSeek-R1DeepSeek · Open weight
40.3%
75
GPT-5.1-Codex-MaxOpenAI · Closed
40.2%
76
GPT-5.1-CodexOpenAI · Closed
40.2%
77
Step 3.7 FlashStepFun · Open weight
40.0%
78
Gemma 4 26B A4BGoogle · Open weight
40.0%
79
Nemotron 3 UltraNVIDIA · Open weight
39.9%
80
o3-miniOpenAI · Closed
39.9%
81
Qwen3.6-27BAlibaba · Open weight
39.8%
82
Mistral Medium 3.5 128BMistral · Open weight
39.6%
83
Qwen3.5-27BAlibaba · Open weight
39.5%
84
GPT-5 miniOpenAI · Closed
39.2%
85
DeepSeek V3.1 (Reasoning)DeepSeek · Open weight
39.1%
86
GPT-OSS 120BOpenAI · Open weight
38.9%
87
DeepSeek V3.2DeepSeek · Open weight
38.7%
88
Mercury 2Inception · Closed
38.7%
89
Step 3.5 FlashStepFun · Open weight
38.5%
90
Qwen3 MaxAlibaba · Closed
38.3%
91
Gemma 4 12BGoogle · Open weight
38.2%
92
GPT-4.1OpenAI · Closed
38.1%
93
Mistral Small 4Mistral · Open weight
38.0%
94
Mistral Small 4 (Reasoning)Mistral · Open weight
38.0%
95
Command A+Cohere · Open weight
37.8%
96
Qwen3.5-35B-A3BAlibaba · Open weight
37.7%
97
DeepSeek R1 Distill Qwen 32BDeepSeek · Open weight
37.6%
98
MiniMax M1 80kMiniMax · Closed
37.4%
99
Claude 4 SonnetAnthropic · Closed
37.3%
100
MiMo-V2-OmniXiaomi · Closed
36.7%
101
DeepSeek V3.1DeepSeek · Open weight
36.7%
102
GPT-5 nanoOpenAI · Closed
36.6%
103
Mistral Large 3Mistral · Closed
36.2%
104
Grok Code Fast 1xAI · Closed
36.2%
105
Trinity-Large-PreviewArcee AI · Open weight
36.1%
106
Trinity-Large-ThinkingArcee AI · Open weight
36.1%
107
Nemotron 3 Super 120B A12BNVIDIA · Open weight
36.0%
108
Qwen3.6-35B-A3BAlibaba · Open weight
35.8%
109
o1OpenAI · Closed
35.8%
110
K-ExaoneLG AI Research · Closed
35.6%
111
DeepSeek V3DeepSeek · Open weight
35.4%
112
Nemotron Ultra 253BNVIDIA · Open weight
34.7%
113
Kimi K2Moonshot AI · Closed
34.5%
114
GPT-OSS 20BOpenAI · Open weight
34.4%
115
GLM-4.7-FlashZ.AI · Open weight
33.7%
116
GPT-4oOpenAI · Closed
33.3%
117
GLM-4.6Z.AI · Open weight
33.1%
118
Llama 4 MaverickMeta · Open weight
33.1%
119
Mistral Medium 3Mistral · Closed
33.1%
120
GPT-4 TurboOpenAI · Closed
31.9%
121
GLM-4.5-AirZ.AI · Closed
30.6%
122
Llama 3.1 405BMeta · Open weight
29.9%
123
Grok 4.1 FastxAI · Closed
29.6%
124
Nemotron 3 Nano 30BNVIDIA · Open weight
29.6%
125
Gemini 1.5 ProGoogle · Closed
29.5%
126
Mistral Large 2Mistral · Closed
29.2%
127
Gemini 2.5 FlashGoogle · Closed
29.1%
128
Nemotron 3 Nano Omni 30B A3BNVIDIA · Open weight
27.8%
129
Ling 2.6 FlashInclusionAI · Open weight
27.1%
130
Qwen2.5 Coder 32B InstructAlibaba · Open weight
27.1%
131
Sarvam 105BSarvam · Open weight
26.4%
132
Phi-4Microsoft · Open weight
26.0%
133
MiMo-V2-FlashXiaomi · Open weight
25.9%
134
GPT-4.1 nanoOpenAI · Closed
25.9%
135
Exaone 4.0 32BLG AI Research · Open weight
25.2%
136
Solar Pro 2Upstage · Closed
24.8%
137
Gemma 4 E4BGoogle · Open weight
24.4%
138
Ministral 3 14B (Reasoning)Mistral · Open weight
23.6%
139
Ministral 3 14BMistral · Open weight
23.6%
140
Claude 3 OpusAnthropic · Closed
23.3%
141
GPT-4o miniOpenAI · Closed
22.9%
142
Gemma 3 27BGoogle · Open weight
21.2%
143
Gemma 4 E2BGoogle · Open weight
20.9%
144
Nova ProAmazon · Closed
20.8%
145
Ministral 3 8B (Reasoning)Mistral · Open weight
20.8%
146
Ministral 3 8BMistral · Open weight
20.8%
147
Sarvam 30BSarvam · Open weight
19.2%
148
Claude 3 HaikuAnthropic · Closed
18.6%
149
Llama 4 ScoutMeta · Open weight
17.0%
150
Ministral 3 3B (Reasoning)Mistral · Open weight
14.4%
151
Ministral 3 3BMistral · Open weight
14.4%
152
Gemini 1.0 ProGoogle · Closed
11.7%
153
LFM2-24B-A2BLiquidAI · Closed
10.9%
154
Granite-4.0-1BIBM · Open weight
8.7%
155
Granite-4.0-H-1BIBM · Open weight
8.2%
156
LFM2.5-8B-A1BLiquidAI · Open weight
7.8%
157
Exaone 4.0 1.2BLG AI Research · Open weight
7.4%
158
LFM2.5-1.2B-ThinkingLiquidAI · Closed
4.2%
159
LFM2.5-VL-1.6B-ExtractLiquidAI · Open weight
3.0%
160
LFM2.5-1.2B-InstructLiquidAI · Closed
2.3%
161
Granite-4.0-H-350MIBM · Open weight
1.7%
162
Granite-4.0-350MIBM · Open weight
0.9%

The published AA-SciCode snapshot places Claude Fable 5 first at 60.2%. The third row is 1.5 points behind. The broader top-10 range is 5.7 points, so many of the published results sit in a relatively narrow band.

162 models have been evaluated on AA-SciCode. The benchmark falls in the Coding category. This category carries a 20% weight in BenchLM.ai's overall scoring system. AA-SciCode is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About AA-SciCode

Year

2026

Tasks

Scientific coding subproblems

Format

Task success rate

Difficulty

Scientific programming

BenchLM stores the Artificial Analysis SciCode result separately from the weighted SciCode lane so AA refreshes remain display-only.

Artificial Analysis SciCode Benchmark Leaderboard

BenchLM freshness & provenance

Version

AA-SciCode 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

FAQ

What does AA-SciCode measure?

A display-only Artificial Analysis SciCode score.

Which model scores highest on AA-SciCode?

Claude Fable 5 by Anthropic currently leads with a score of 60.2% on AA-SciCode.

How many models are evaluated on AA-SciCode?

162 AI models have been evaluated on AA-SciCode on BenchLM.

Compare Top Models on AA-SciCode

Claude Fable 5 vs Gemini 3.1 ProGemini 3.1 Pro vs Kimi K3Kimi K3 vs Muse Spark 1.1Muse Spark 1.1 vs GPT-5.4
Last updated: July 23, 2026 · BenchLM version AA-SciCode 2026

Choose a model with this week’s evidence

Join 2,000+ readers for ranking moves, pricing changes, and the claims that still need proof.

One email each week. Unsubscribe anytime.