Coding benchmark

SciCode leaderboard

As of October 2026, Claude Opus 5.5 has the highest published SciCode score on Noometry at 66.9%, out of 121 models with results.

Last verified October 10, 2026

About SciCode

Scientific research coding problems from physics, chemistry, biology and math, broken into sub-steps.

Category: Coding
Introduced: 2024
Format: Code generation
Unit: Percent (random guessing ≈ 0%)
Official site: scicode-bench.github.io

Top 15 models

Claude Opus 5.5 66.9%
Claude Fable 5.1 63.1%
Gemini 4 Argon 61.8%
Claude Fable 5 61%
Claude Sonnet 5.5 61%
MiMo-V2.6-Pro 60.9%
Gemini 3.7 Flash 59.8%
Muse Spark 1.3 59.7%
Kimi K3 59.5%
GLM-5.3 59%
Gemini 3.1 Pro Preview 58.9%
Step 5 Preview 58.9%
Muse Spark 1.1 58.8%
Grok 4.7 57.8%
GPT-6 Sol 57.6%
5055606570

Sponsored placements are available on pages like this one. Advertise on Noometry

All results

SciCode results by model
#	Model	Provider	Score	Setting	Source
1	Claude Opus 5.5	Anthropic	66.9%	max	Epoch AI
2	Claude Fable 5.1	Anthropic	63.1%	max	Epoch AI
3	Gemini 4 Argon	Google	61.8%	high	Epoch AI
4	Claude Fable 5	Anthropic	61%	max	Epoch AI
5	Claude Sonnet 5.5	Anthropic	61%	max	Epoch AI
6	MiMo-V2.6-Pro	Xiaomi	60.9%		Epoch AI
7	Gemini 3.7 Flash	Google	59.8%	medium	Epoch AI
8	Muse Spark 1.3	Meta	59.7%	xhigh	Epoch AI
9	Kimi K3	Moonshot AI	59.5%	max	Epoch AI
10	GLM-5.3	Z.ai (Zhipu)	59%	max	Epoch AI
11	Gemini 3.1 Pro Preview	Google	58.9%		Epoch AI
12	Step 5 Preview	StepFun	58.9%		Epoch AI
13	Muse Spark 1.1	Meta	58.8%		Epoch AI
14	Grok 4.7	xAI	57.8%	high	Epoch AI
15	GPT-6 Sol	OpenAI	57.6%	max	Epoch AI
16	GPT-5.6 Sol	OpenAI	57.1%	max	Epoch AI
17	Gemini 3.8 Flash	Google	56.6%	high	Epoch AI
18	GPT-5.4	OpenAI	56.6%	xhigh	Epoch AI
19	GPT-6 Astra	OpenAI	56.5%	max	Epoch AI
20	Grok 4.6	xAI	56.5%	high	Epoch AI
21	Claude Opus 5	Anthropic	56.4%	max	Epoch AI
22	Muse Spark 1.2	Meta	56.4%	xhigh	Epoch AI
23	GPT-5.5	OpenAI	56.1%	xhigh	Epoch AI
24	GPT-6.1 Sol	OpenAI	55.8%	high	Epoch AI
25	GPT-5.6 Terra	OpenAI	55%	max	Epoch AI
26	GPT-6 Luna	OpenAI	54.6%	max	Epoch AI
27	Claude Opus 4.7	Anthropic	54.5%	max	Epoch AI
28	Claude Sonnet 5	Anthropic	54.3%	high	Epoch AI
29	Grok 4.5	xAI	54.1%	high	Epoch AI
30	GPT-5.6 Luna	OpenAI	53.6%	max	Epoch AI
31	Claude Opus 4.8	Anthropic	53.5%	max	Epoch AI
32	Kimi K2.6	Moonshot AI	53.5%		Epoch AI
33	Qwen3.8 Max	Alibaba (Qwen)	53.2%		Epoch AI
34	Gemini 3.5 Flash	Google	53.1%	high	Epoch AI
35	Gemini 3.6 Flash	Google	52.7%	high	Epoch AI
36	DeepSeek V4.1 Flash	DeepSeek	51.9%	max	Epoch AI
37	GLM-5.3-Flash	Z.ai (Zhipu)	51.6%		Epoch AI
38	Muse Spark	Meta	51.5%		Epoch AI
39	MiMo-V2.6-Flash	Xiaomi	51.3%		Epoch AI
40	DeepSeek V4 Pro	DeepSeek	51%	max	Epoch AI
41	GLM-5.2	Z.ai (Zhipu)	50.5%	max	Epoch AI
42	Grok Build 0.1	xAI	50.2%		Epoch AI
43	MiMo-V2.5-Pro	Xiaomi	50.2%		Epoch AI
44	DeepSeek V4 Flash	DeepSeek	49.9%	max	Epoch AI
45	GPT-5.4 mini	OpenAI	49.9%	xhigh	Epoch AI
46	Kimi K2.5	Moonshot AI	49%		Epoch AI
47	Qwen3.7 Max	Alibaba (Qwen)	48.8%		Epoch AI
48	Inkling-Small	Thinking Machines Lab	48.7%		Epoch AI
49	GPT-5.5 Instant	OpenAI	48.6%		Epoch AI
50	Kimi K2.7 Code	Moonshot AI	47.5%		Epoch AI
51	Grok 4.3	xAI	47.3%	high	Epoch AI
52	MiniMax-M3	MiniMax	47.1%		Epoch AI
53	Inkling	Thinking Machines Lab	47%	xhigh	Epoch AI
54	MiniMax-M2.7	MiniMax	47%		Epoch AI
55	GPT-5.4 nano	OpenAI	46.9%	xhigh	Epoch AI
56	Claude Sonnet 4.6	Anthropic	46.8%	max	Epoch AI
57	Qwen3.8 27B	Alibaba (Qwen)	46.6%	xhigh	Epoch AI
58	Qwen3.7 Plus	Alibaba (Qwen)	45.5%		Epoch AI
59	GLM-4.7	Z.ai (Zhipu)	45.1%		Epoch AI
60	Muse Glimmer	Meta	44.9%	high	Epoch AI
61	Claude Sonnet 4.5	Anthropic	44.7%		Epoch AI
62	GLM-5.1	Z.ai (Zhipu)	43.8%		Epoch AI
63	Gemma 4 31B IT	Google	43.4%		Epoch AI
64	Claude Haiku 4.5	Anthropic	43.3%		Epoch AI
65	GPT-5.1	OpenAI	43.3%		Epoch AI
66	MiMo-V2.5	Xiaomi	43.1%		Epoch AI
67	GPT-5	OpenAI	42.9%		Epoch AI
68	Gemini 2.5 Pro	Google	42.8%		Epoch AI
69	Nova 2.0 Pro Preview	Amazon	42.7%	medium	Epoch AI
70	Qwen3 235B-A22B	Alibaba (Qwen)	42.4%		Epoch AI
71	Ring-2.6-1T	Ant Group (inclusionAI)	42.4%		Epoch AI
72	Gemini 3.1 Flash Lite	Google	41.9%		Epoch AI
73	Gemini 3.5 Flash Lite	Google	41.3%		Epoch AI
74	Qwen3.6 Plus	Alibaba (Qwen)	40.7%		Epoch AI
75	DeepSeek-V3.1-Terminus	DeepSeek	40.6%		Epoch AI
76	GPT-4.1 mini	OpenAI	40.4%		Epoch AI
77	Nemotron 3 Ultra	NVIDIA	40.3%		Epoch AI
78	Mistral Medium	Mistral AI	40.2%		Epoch AI
79	Claude Sonnet 4	Anthropic	40%		Epoch AI
80	Gemma 4 26B A4B IT	Google	40%		Epoch AI
81	Step 3.7 Flash	StepFun	40%		Epoch AI
82	o3-mini	OpenAI	39.8%	high	Epoch AI
83	GPT-5 Mini	OpenAI	39.2%		Epoch AI
84	Magistral Medium	Mistral AI	39.2%		Epoch AI
85	DeepSeek-V3.2-Exp	DeepSeek	38.9%	thinking	Epoch AI
86	Mercury 2	Inception	38.7%		Epoch AI
87	Mercury 2.5	Inception	38.5%		Epoch AI
88	GLM-4.6	Z.ai (Zhipu)	38.4%		Epoch AI
89	Command A Plus	Cohere	37.8%		Epoch AI
90	Qwen3.6 27B	Alibaba (Qwen)	37.3%	none	Epoch AI
91	Mistral Large	Mistral AI	36.2%		Epoch AI
92	Trinity Large Thinking	Arcee AI	36.1%		Epoch AI
93	gpt-oss-120b	OpenAI	36%	low	Epoch AI
94	Nemotron 3 Super	NVIDIA	36%		Epoch AI
95	DeepSeek-V3	DeepSeek	35.8%		Epoch AI
96	Qwen3.6 35B-A3B	Alibaba (Qwen)	35.8%		Epoch AI
97	DeepSeek-R1	DeepSeek	35.7%		Epoch AI
98	Qwen3.5 122B-A10B	Alibaba (Qwen)	35.6%	none	Epoch AI
99	Qwen3 32B	Alibaba (Qwen)	35.4%		Epoch AI
100	Magistral Small	Mistral AI	35.2%		Epoch AI
101	gpt-oss-20b	OpenAI	34.4%	high	Epoch AI
102	Qwen3-30B-A3B	Alibaba (Qwen)	33.3%		Epoch AI
103	Llama 4 Maverick	Meta	33.1%		Epoch AI
104	Qwen3 Coder Next	Alibaba (Qwen)	32.3%		Epoch AI
105	Qwen3 14B	Alibaba (Qwen)	31.6%		Epoch AI
106	Qwen3.5 35B-A3B	Alibaba (Qwen)	29.3%	none	Epoch AI
107	Devstral Small 2505	Mistral AI	28.8%		Epoch AI
108	Qwen3.5-9B	Alibaba (Qwen)	27.5%		Epoch AI
109	Claude 3.5 Haiku	Anthropic	27.4%		Epoch AI
110	Mistral Small	Mistral AI	26.5%		Epoch AI
111	Llama-3.3-70B-Instruct	Meta	26%		Epoch AI
112	GPT-4.1 nano	OpenAI	25.9%		Epoch AI
113	MiMo-V2-Flash	Xiaomi	25.9%		Epoch AI
114	Granite 4.1 30B	IBM	25.8%		Epoch AI
115	Solar Pro 3	Upstage	24.7%		Epoch AI
116	Qwen3 8B	Alibaba (Qwen)	22.6%		Epoch AI
117	Gemma 3 27B	Google	21.2%		Epoch AI
118	Gemma 3 12B	Google	17.4%		Epoch AI
119	Llama 4 Scout	Meta	17%		Epoch AI
120	Llama 3.1-8B	Meta	13.2%		Epoch AI
121	Phi-4 Mini	Microsoft	10.8%		Epoch AI

Compare the leaders

Other coding benchmarks

Frequently asked questions

What does SciCode measure?

Scientific research coding problems from physics, chemistry, biology and math, broken into sub-steps.

Which model has the highest SciCode score?

As of October 2026, Claude Opus 5.5 has the highest published SciCode score on Noometry at 66.9%, out of 121 models with results.

What is the best open-weight model on SciCode?

MiMo-V2.6-Pro has the highest SciCode accuracy among open-weight models at 60.9%, ranking 6 of 121 overall.