llmrun Composite Score · Science

The best open LLMs for science, ranked on graduate-level science and knowledge boards such as GPQA. The score shown is the overall llmrun Composite Score; the highlighted column is the scientific knowledge skill score used for the order.

45 benchmarks4 sourcesUpdated 17 Oct 2026Reference scale 2026-Q4How the Composite Score is built

Science score vs VRAM needed

This chart doesn't show proprietary models.

Open models Overall

General Coding Agents Math Science Reasoning

Models ranked by science llmrun Composite Score
# (open / overall)	Model	llmrun Composite Score	Coding	Agents	Math	Science	Reasoning	Benchmarks	VRAM
1 / 14	MiMo V2.6 Pro RL614 GB · 8 benchmarks		126	132	–	140	–	8	614 GB
2 / 28	GLM 5.3454 GB · 26 benchmarks		130	133	129	131	130	26	454 GB
3 / 34	GLM 5.2454 GB · 26 benchmarks		120	128	126	129	120	26	454 GB
4 / 44	DeepSeek V4 Flash 0731183 GB · 21 benchmarks		121	–	128	124	129	21	183 GB
5 / 46	GLM 5.3 Flash205 GB · 24 benchmarks		117	127	121	124	127	24	205 GB
6 / 47	DeepSeek V4.1 Flash458 GB · 24 benchmarks		122	128	129	123	134	24	458 GB
7 / 50	MiMo V2.6 Flash RL187 GB · 4 benchmarks		124	–	–	122	–	4	187 GB
8 / 51	Kimi K2.6618 GB · 20 benchmarks		119	122	122	122	118	20	618 GB
9 / 55	Inkling Small163 GB · 11 benchmarks		–	–	117	118	123	11	163 GB
10 / 56	MiniMax M3258 GB · 22 benchmarks		115	112	106	118	108	22	258 GB
11 / 58	Kimi K2.7 Code618 GB · 19 benchmarks		115	119	121	118	115	19	618 GB
12 / 60	GLM 5.1454 GB · 14 benchmarks		115	123	117	117	116	14	454 GB
13 / 62	Kimi K2.5618 GB · 19 benchmarks		110	111	–	117	115	19	618 GB
14 / 63	Inkling575 GB · 25 benchmarks		110	111	112	116	121	25	575 GB
15 / 65	MiMo V2.5 Pro616 GB · 6 benchmarks		116	–	–	115	111	6	616 GB
16 / 70	MiniMax M2.7140 GB · 7 benchmarks		103	112	–	112	–	7	140 GB
17 / 72	Qwen3.8 27B17.9 GB · 15 benchmarks		120	123	118	112	124	15	17.9 GB
18 / 73	NVIDIA Nemotron 3 Ultra 550B A55B BF16378 GB · 17 benchmarks		104	102	112	112	116	17	378 GB
19 / 75	DeepSeek V4 Flash175 GB · 10 benchmarks		107	–	–	111	112	10	175 GB
20 / 76	GLM 4.7218 GB · 13 benchmarks		101	106	111	111	111	13	218 GB
21 / 77	Qwen3.6 27B17.9 GB · 11 benchmarks		102	–	114	111	105	11	17.9 GB
22 / 78	Qwen3.6 35B A3B22.0 GB · 11 benchmarks		105	–	109	110	104	11	22.0 GB
23 / 81	Muse Glimmer 30B19.6 GB · 2 benchmarks	—	–	–	–	108	–	2	19.6 GB
24 / 83	Qwen3.5 35B A3B22.0 GB · 7 benchmarks		–	–	–	108	107	7	22.0 GB
25 / 85	Qwen3 235B A22B Thinking 2507143 GB · 12 benchmarks		102	–	–	108	106	12	143 GB
26 / 86	MiMo V2.5188 GB · 4 benchmarks		103	–	–	107	–	4	188 GB
27 / 89	Gemma 4 31B IT21.7 GB · 13 benchmarks		108	–	–	105	110	13	21.7 GB
28 / 90	Qwen3.5 9B6.6 GB · 8 benchmarks		–	–	–	105	97	8	6.6 GB
29 / 92	GPT OSS 120B70.8 GB · 19 benchmarks		98	82	–	103	96	19	70.8 GB
30 / 94	Step 3.7 Flash139 GB · 3 benchmarks	—	107	–	–	102	–	3	139 GB
31 / 95	Cogito 671B V2.1404 GB · 2 benchmarks	—	–	–	–	102	–	2	404 GB
32 / 96	Gemma 4 26B A4B IT16.5 GB · 8 benchmarks		104	–	–	102	102	8	16.5 GB
33 / 98	DeepSeek V3.2 Exp413 GB · 10 benchmarks		107	–	–	101	113	10	413 GB
34 / 101	GLM 4.6217 GB · 6 benchmarks		96	104	–	100	–	6	217 GB
35 / 102	Command A Plus 05 2026 BF16132 GB · 2 benchmarks	—	–	–	–	100	–	2	132 GB
36 / 103	DeepSeek V3.1 Terminus412 GB · 5 benchmarks		107	–	–	99	108	5	412 GB
37 / 105	DeepSeek R1412 GB · 13 benchmarks		100	–	–	99	96	13	412 GB
38 / 106	Qwen3 30B A3B Thinking 250719.4 GB · 7 benchmarks		–	–	–	99	94	7	19.4 GB
39 / 107	DeepSeek v3 0324412 GB · 12 benchmarks		100	–	–	98	88	12	412 GB
40 / 110	NVIDIA Nemotron 3 Super 120B A12B BF1674.7 GB · 4 benchmarks		96	–	–	98	–	4	74.7 GB
41 / 112	Mistral Large 3 675B Instruct 2512454 GB · 5 benchmarks		93	–	–	97	89	5	454 GB
42 / 113	Trinity Large Thinking271 GB · 3 benchmarks	—	92	–	–	97	–	3	271 GB
43 / 116	Qwen3 32B22.6 GB · 8 benchmarks		93	–	–	96	92	8	22.6 GB
44 / 117	Qwen3.5 122B A10B75.5 GB · 6 benchmarks		–	–	–	96	112	6	75.5 GB
45 / 118	Llama 4 Maverick 17B 128E Instruct273 GB · 16 benchmarks		81	–	–	96	85	16	273 GB
46 / 120	Qwen3 14B11.9 GB · 7 benchmarks		–	–	–	94	88	7	11.9 GB
47 / 121	GPT OSS 20B13.1 GB · 10 benchmarks		101	–	–	92	91	10	13.1 GB
48 / 123	Qwen3 Coder Next48.3 GB · 3 benchmarks	—	95	–	–	91	–	3	48.3 GB
49 / 125	DeepSeek v3412 GB · 8 benchmarks		97	–	–	90	81	8	412 GB
50 / 129	Qwen3 8B7.6 GB · 7 benchmarks		–	–	–	88	81	7	7.6 GB
51 / 130	Devstral Small 2 24B Instruct 251218.1 GB · 3 benchmarks	—	–	–	–	86	–	3	18.1 GB
52 / 131	Llama 4 Scout 17B 16E Instruct77.0 GB · 11 benchmarks		–	–	–	83	80	11	77.0 GB
53 / 132	Magistral Small 250918.1 GB · 5 benchmarks		–	–	–	83	–	5	18.1 GB
54 / 133	Mistral Small 3.2 24B Instruct 250618.1 GB · 5 benchmarks		–	–	–	82	–	5	18.1 GB
55 / 134	MiMo v2 Flash187 GB · 3 benchmarks	—	100	–	–	82	–	3	187 GB
56 / 136	Granite 4.1 30B21.9 GB · 2 benchmarks	—	–	–	–	82	–	2	21.9 GB
57 / 139	Mistral Small 3.1 24B Instruct 250318.1 GB · 5 benchmarks		–	–	–	81	–	5	18.1 GB
58 / 140	Llama 3.3 70B Instruct51.3 GB · 12 benchmarks		76	–	–	81	82	12	51.3 GB
59 / 141	Gemma 3 27B IT21.7 GB · 9 benchmarks		70	–	–	81	74	9	21.7 GB
60 / 143	Gemma 3 12B IT10.9 GB · 7 benchmarks		–	–	–	71	66	7	10.9 GB
61 / 144	Phi 4 Mini Instruct4.8 GB · 3 benchmarks	—	–	–	–	61	–	3	4.8 GB
62 / 145	Llama 3.1 8B Instruct7.9 GB · 9 benchmarks		56	–	–	56	69	9	7.9 GB

The range after each score is a 90% interval: given the boards a model has results on, its true score is very likely inside it. Short ranges mean many agreeing results; long ranges mean few or conflicting ones.

Why some models are missing. A model receives a score only if it has results on at least 4 benchmarks across at least 2 categories, with one being reasoning or coding, or on 3 benchmarks that narrow its 90% range to at most 15.6 points. The results must come from at least two different benchmarks: categories of one benchmark, such as LiveBench, count once. Skill scores require at least 2 benchmarks for that skill; otherwise they display as –.

VRAM is llmrun's estimate at Q4_K_M (or the smallest quantization we track) for the weights plus room for a 16K-token context. Longer contexts need more memory. Proprietary models can't run locally, so the VRAM filter hides them.

How the llmrun Composite Score is built · llmrun doesn't run these benchmarks; scores come from public sources.