llmrun Composite Score · Coding

The best open LLMs for coding, ranked on SWE-bench, LiveCodeBench, Aider and other coding boards. The score shown is the overall llmrun Composite Score; the highlighted column is the coding skill score used for the order.

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

Coding score vs VRAM needed

This chart doesn't show proprietary models.

Open models Overall

General Coding Agents Math Science Reasoning

Models ranked by coding llmrun Composite Score
# (open / overall)	Model	llmrun Composite Score	Coding	Agents	Math	Science	Reasoning	Benchmarks	VRAM
1 / 34	MiMo V2.6 Flash RL187 GB · 4 benchmarks		124	–	–	122	–	4	187 GB
2 / 38	DeepSeek V4 Flash 0731183 GB · 21 benchmarks		121	–	128	124	129	21	183 GB
3 / 42	Qwen3.8 27B17.9 GB · 15 benchmarks		120	123	118	112	124	15	17.9 GB
4 / 45	GLM 5.3 Flash205 GB · 24 benchmarks		117	127	121	124	127	24	205 GB
5 / 55	Smaug Flash208 GB · 6 benchmarks	—	113	–	–	–	124	6	208 GB
6 / 65	MiniMax M2.5140 GB · 11 benchmarks		109	110	–	–	114	11	140 GB
7 / 67	Gemma 4 31B IT21.7 GB · 13 benchmarks		108	–	–	105	110	13	21.7 GB
8 / 71	Step 3.7 Flash139 GB · 3 benchmarks	—	107	–	–	102	–	3	139 GB
9 / 72	DeepSeek V4 Flash175 GB · 10 benchmarks		107	–	–	111	112	10	175 GB
10 / 74	Qwen3.6 35B A3B22.0 GB · 11 benchmarks		105	–	109	110	104	11	22.0 GB
11 / 76	Gemma 4 26B A4B IT16.5 GB · 8 benchmarks		104	–	–	102	102	8	16.5 GB
12 / 81	MiMo V2.5188 GB · 4 benchmarks		103	–	–	107	–	4	188 GB
13 / 83	MiniMax M2.7140 GB · 7 benchmarks		103	112	–	112	–	7	140 GB
14 / 84	Qwen3 235B A22B Thinking 2507143 GB · 12 benchmarks		102	–	–	108	106	12	143 GB
15 / 85	Qwen3.6 27B17.9 GB · 11 benchmarks		102	–	114	111	105	11	17.9 GB
16 / 89	Qwen3 235B A22B Instruct 2507143 GB · 8 benchmarks		102	–	–	–	98	8	143 GB
17 / 91	GLM 4.7218 GB · 13 benchmarks		101	106	111	111	111	13	218 GB
18 / 92	Qwen3 235B A22B143 GB · 9 benchmarks		101	–	–	–	99	9	143 GB
19 / 93	DeepSeek V4 Flash Vision Exp183 GB · 6 benchmarks	—	101	–	–	–	123	6	183 GB
20 / 94	GPT OSS 20B13.1 GB · 10 benchmarks		101	–	–	92	91	10	13.1 GB
21 / 96	MiMo v2 Flash187 GB · 3 benchmarks	—	100	–	–	82	–	3	187 GB
22 / 102	GLM 4.5218 GB · 7 benchmarks		99	105	–	–	–	7	218 GB
23 / 105	Qwen3.5 27B17.9 GB · 6 benchmarks		98	–	–	–	111	6	17.9 GB
24 / 106	GPT OSS 120B70.8 GB · 19 benchmarks		98	82	–	103	96	19	70.8 GB
25 / 109	GLM 4.6217 GB · 6 benchmarks		96	104	–	100	–	6	217 GB
26 / 110	NVIDIA Nemotron 3 Super 120B A12B BF1674.7 GB · 4 benchmarks		96	–	–	98	–	4	74.7 GB
27 / 112	Qwen3 Coder Next48.3 GB · 3 benchmarks	—	95	–	–	91	–	3	48.3 GB
28 / 113	Qwen3 32B22.6 GB · 8 benchmarks		93	–	–	96	92	8	22.6 GB
29 / 121	Llama 3.3 70B Instruct51.3 GB · 12 benchmarks		76	–	–	81	82	12	51.3 GB
30 / 123	Gemma 3 27B IT21.7 GB · 9 benchmarks		70	–	–	81	74	9	21.7 GB
31 / 125	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.