llmrun Composite Score · Agents

The best open LLMs for agents, ranked on agentic and tool-use boards such as Terminal-Bench and tau-bench. The score shown is the overall llmrun Composite Score; the highlighted column is the agentic tool use skill score used for the order.

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

Agents score vs VRAM needed

This chart doesn't show proprietary models.

Open models Overall

General Coding Agents Math Science Reasoning

Models ranked by agents llmrun Composite Score
# (open / overall)	Model	Coding	Agents	Math	Science	Reasoning	Benchmarks	VRAM
1 / 23	GLM 5.3454 GB · 26 benchmarks	130	133	129	131	130	26	454 GB
2 / 33	GLM 5.2454 GB · 26 benchmarks	120	128	126	129	120	26	454 GB
3 / 37	DeepSeek V4.1 Flash458 GB · 24 benchmarks	122	128	129	123	134	24	458 GB
4 / 38	GLM 5.3 Flash205 GB · 24 benchmarks	117	127	121	124	127	24	205 GB
5 / 43	Qwen3.8 27B17.9 GB · 15 benchmarks	120	123	118	112	124	15	17.9 GB
6 / 44	GLM 5.1454 GB · 14 benchmarks	115	123	117	117	116	14	454 GB
7 / 49	GLM 5454 GB · 14 benchmarks	110	118	–	–	109	14	454 GB
8 / 55	MiniMax M3258 GB · 22 benchmarks	115	112	106	118	108	22	258 GB
9 / 57	MiniMax M2.7140 GB · 7 benchmarks	103	112	–	112	–	7	140 GB
10 / 60	MiniMax M2.5140 GB · 11 benchmarks	109	110	–	–	114	11	140 GB
11 / 63	DeepSeek V3.2413 GB · 13 benchmarks	–	108	–	–	112	13	413 GB
12 / 65	GLM 4.7218 GB · 13 benchmarks	101	106	111	111	111	13	218 GB
13 / 66	GLM 4.5218 GB · 7 benchmarks	99	105	–	–	–	7	218 GB
14 / 68	Qwen3 Coder 480B A35B Instruct291 GB · 6 benchmarks	101	105	–	–	–	6	291 GB
15 / 69	MiniMax M2140 GB · 5 benchmarks	–	105	–	–	–	5	140 GB
16 / 70	GLM 4.6217 GB · 6 benchmarks	96	104	–	100	–	6	217 GB
17 / 73	NVIDIA Nemotron 3 Ultra 550B A55B BF16378 GB · 17 benchmarks	104	102	112	112	116	17	378 GB
18 / 81	Qwen2.5 Coder 32B Instruct24.3 GB · 3 benchmarks	–	84	–	–	–	3	24.3 GB
19 / 82	GPT OSS 120B70.8 GB · 19 benchmarks	98	82	–	103	96	19	70.8 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.