Agent & Tool-Use Benchmarks

Which AI models handle function calling, MCP tool use, browsing, and multi-step agent workflows best? Verified-ranked results across 24 agentic benchmarks.

Agentic carries 22% weightin BenchLM.ai's overall score — the single biggest category.

This page now shows only core agentic benchmark rows with an attached exact source record. Source-unverified manual rows are excluded from the displayed agentic score and table cells.

Best Agentic Model

GPT-5.6 Sol

Verified score: 92 · OpenAI

Best Open-Weight Agent

Qwen3.8 Max

Verified score: 86.1 · Alibaba

Benchmarks Tracked

26 benchmarks

Terminal, browsing, tool-use, and computer-use

Benchmark Categories

Core Weighted (3)

These 3 benchmarks determine agentic rankings

Terminal-Bench 2.0 BrowseComp OSWorld-Verified

Tool Calling & MCP (6)

Function calling, MCP tool use, and structured workflows

MCP Atlas MCP-Tasks Toolathlon BFCL v4 τ²-bench results τ³-bench results

Agent Frameworks (8)

OpenClaw-style and end-to-end agent evaluations

ZClawBench QwenClawBench QwenWebBench MM-ClawBench Claw-Eval PinchBench OpenHands Index SWE-Atlas Refactoring

Computer & Browser Use (5)

Desktop GUI, mobile, and browser navigation tasks

OSWorld AndroidWorld WebVoyager BrowseComp-VL DeepPlanning

Specialized (4)

Domain-specific agentic tasks across ML, research, and airline

MLE-Bench Lite VITA-Bench WideResearch τ²-bench Airline

Top 15 Models by Weighted Agentic Score

OpenAIAnthropicGoogleMetaDeepSeekMistralxAIAlibaba

78 models match these filters.

CSV JSON

#	Model
1	GPT-5.6 Sol OpenAI · Proprietary	92	91.9	92.2	—	82
2	Claude Opus 5 Anthropic · Proprietary	90.8	—	90.8	—	83.07
3	GPT-5.5 Pro OpenAI · Proprietary	90.1	—	90.1	—	64.17
4	Kimi K3 Moonshot AI · Pending	89.5	88.3	91.2	—	80.5
5	GPT-5.4 Pro OpenAI · Proprietary	89.3	—	89.3	—	61.26
6	GPT-5.6 Terra OpenAI · Proprietary	87.4	87.4	87.5	—	72.91
7	Claude Mythos 5 Anthropic · Proprietary	87	88	88	85	83.21
8	Qwen3.8 Max Alibaba · Open Weight	86.1	—	—	86.1	79.91
9	Claude Fable 5 Anthropic · Proprietary	84.6	84.3	—	85	82.96
10	Qwen3.8-27B Alibaba · Open Weight	84.3	—	—	84.3	—
11	GPT-5.6 Luna OpenAI · Proprietary	84.1	84.7	83.3	—	67.29
12	dots3-note Preview Dots Studio · Open Weight	83.3	—	83.3	—	—
13	Grok 4.5 xAI · Proprietary	83.3	83.3	—	—	75.42
14	Gemini 3.6 Flash Google · Proprietary	83	—	—	83	75.54
15	Holo3-35B-A3B H Company · Open Weight	82.6	—	—	82.56	—
16	Sakana Fugu-Ultra Sakana AI · Proprietary	82.1	82.1	—	—	—
17	Claude Sonnet 5 Anthropic · Proprietary	81.9	80.4	84.7	81.2	64.78
18	GPT-5.5 OpenAI · Proprietary	81.6	82	84.4	78.7	73.37
19	SWE-1.7 Cognition · Proprietary	81.5	81.5	—	—	—
20	GLM-5.2 Z.AI · Open Weight	81	81	—	—	63.34
21	Muse Spark 1.1 Meta · Proprietary	80.4	80	—	80.8	76.88
22	Claude Opus 4.8 Anthropic · Proprietary	80.3	74.6	84.3	83.4	76.58
23	Sakana Fugu Sakana AI · Proprietary	80.2	80.2	—	—	—
24	Holo3-122B-A10B H Company · Proprietary	78.9	—	—	78.85	—
25	Ornith-1.0-397B DeepReinforce AI · Open Weight	77.5	77.5	—	—	—
26	Gemini 3.5 Flash Google · Proprietary	77.2	76.2	—	78.4	64.67
27	GPT-5.4 OpenAI · Proprietary	77.2	75.1	82.7	75	73.36
28	Agents-A1 InternScience · Open Weight	75.5	—	75.51	—	60.79
29	Claude Opus 4.7 (Adaptive) Anthropic · Proprietary	75.1	69.4	79.3	78	72.13
30	DeepSeek V4 Pro 0813 DeepSeek · Proprietary	74.5	67.9	83.4	—	61.16
31	Kimi K2.6 Moonshot AI · Open Weight	73.5	66.7	83.2	73.1	60.21
32	Claude Opus 4.6 Anthropic · Proprietary	73	65.4	83.7	72.7	67.98
33	MiniMax M3 MiniMax · Open Weight	72.3	66	83.52	70.06	68.74
34	Ling 3.0 Flash InclusionAI · Open Weight	72.2	—	72.2	—	53.78
35	Qwen3.7 Plus Alibaba · Proprietary	71.7	70.3	—	73.3	66.02
36	GPT-5.3 Codex OpenAI · Proprietary	71.4	77.3	—	64.7	65.93
37	Laguna S 2.1 Poolside · Open Weight	70.2	70.2	—	—	—
38	Inkling-Small Thinking Machines Lab · Open Weight	70.1	64.7	77.4	—	65.54
39	Qwen3.7 Max Alibaba · Proprietary	69.7	69.7	—	—	71.6
40	Inkling Thinking Machines Lab · Open Weight	69.4	63.8	77.1	—	67.04
41	Composer 2.5 Cursor · Proprietary	69.3	69.3	—	—	—
42	MiMo-V2.5-Pro Xiaomi · Proprietary	68.4	68.4	—	—	69.21
43	Step 3.7 Flash StepFun · Open Weight	66.4	59.5	75.82	—	50.91
44	Muse Glimmer 30B Meta · Open Weight	65.9	—	—	65.9	52.5
45	MiMo-V2.5 Xiaomi · Proprietary	65.8	65.8	—	—	59.16
46	GPT-5.4 mini OpenAI · Proprietary	65.7	60	—	72.1	56.86
47	GLM-5.1 Z.AI · Open Weight	65.4	63.5	68	—	66.91
48	Qwen 3.6 Max (preview) Alibaba · Proprietary	65.4	65.4	—	—	59.99
49	Claude Sonnet 4.6 Anthropic · Proprietary	65.2	59.1	—	72.1	64.53
50	Ornith-1.0-35B DeepReinforce AI · Open Weight	64.2	64.2	—	—	—
51	DeepSeek V4 Flash 0731 DeepSeek · Proprietary	63.8	56.9	73.2	—	—
52	Gemini 3.5 Flash-Lite Google · Proprietary	63.4	54	—	74	65.29
53	Claude Opus 4.5 Anthropic · Proprietary	62.6	59.3	—	66.3	63.63
54	Composer 2 Cursor · Proprietary	61.7	61.7	—	—	—
55	Qwen3.6 Plus Alibaba · Proprietary	61.6	61.6	—	—	64.79
56	Qwen3.6-27B Alibaba · Open Weight	59.3	59.3	—	—	53.73
57	Muse Spark Meta · Proprietary	59	59	—	—	70.81
58	MiniMax M2.7 MiniMax · Open Weight	57	57	—	—	63.06
59	Qwen3.5 397B Alibaba · Open Weight	56.5	52.5	62	—	57.46
60	Qwen3.5-122B-A10B Alibaba · Open Weight	56.4	49.4	63.8	58	59.64
61	GLM-5 Z.AI · Open Weight	56.2	56.2	—	—	65.61
62	GPT-5.2 OpenAI · Proprietary	55.7	—	65.8	47.3	58.24
63	Claude Sonnet 4.5 Anthropic · Proprietary	55.4	50	—	61.4	54.16
64	Kimi K2.5 Moonshot AI · Open Weight	55	50.8	60.6	—	58.89
65	Kimi K2.5 (Reasoning) Moonshot AI · Proprietary	55	50.8	60.6	—	59.81
66	Hy3 Preview Tencent · Open Weight	54.4	54.4	—	—	43.96
67	Qwen3.5-27B Alibaba · Open Weight	52	41.6	61	56.2	59.82
68	Qwen3.6-35B-A3B Alibaba · Open Weight	51.5	51.5	—	—	51.42
69	Nemotron 3 Ultra NVIDIA · Open Weight	51.3	56.4	44.4	—	46.67
70	Qwen3.5-35B-A3B Alibaba · Open Weight	51	40.5	61	54.5	56.15
71	Grok 4.20 xAI · Proprietary	47.1	47.1	—	—	55.19
72	MAI-Thinking-1 Microsoft · Proprietary	46	46	—	—	50.93
73	Laguna M.1 Poolside · Proprietary	45.8	45.8	—	—	—
74	GLM-4.7 Z.AI · Open Weight	45.7	41	52	—	60.66
75	Ornith-1.0-9B DeepReinforce AI · Open Weight	43.1	43.1	—	—	—
76	GPT-5.4 nano OpenAI · Proprietary	42.9	46.3	—	39	66.59
77	Laguna XS.2 Poolside · Open Weight	35.7	35.7	—	—	—
78	Nemotron 3.5 Lightning 30B A3B NVFP4 NVIDIA · Open Weight	29.1	23.46	36.81	—	26.79

RankModelVerified score

GPT-5.6 SolOpenAI · proprietaryTB 91.9BC 92.2OS —

92agentic / 100

Claude Opus 5Anthropic · proprietaryTB —BC 90.8OS —

90.8agentic / 100

GPT-5.5 ProOpenAI · proprietaryTB —BC 90.1OS —

90.1agentic / 100

Kimi K3Moonshot AI · proprietaryTB 88.3BC 91.2OS —

89.5agentic / 100

GPT-5.4 ProOpenAI · proprietaryTB —BC 89.3OS —

89.3agentic / 100

GPT-5.6 TerraOpenAI · proprietaryTB 87.4BC 87.5OS —

87.4agentic / 100

Claude Mythos 5Anthropic · proprietaryTB 88BC 88OS 85

87agentic / 100

Qwen3.8 MaxAlibaba · open weightTB —BC —OS 86.1

86.1agentic / 100

Claude Fable 5Anthropic · proprietaryTB 84.3BC —OS 85

84.6agentic / 100

Qwen3.8-27BAlibaba · open weightTB —BC —OS 84.3

84.3agentic / 100

GPT-5.6 LunaOpenAI · proprietaryTB 84.7BC 83.3OS —

84.1agentic / 100

dots3-note PreviewDots Studio · open weightTB —BC 83.3OS —

83.3agentic / 100

Grok 4.5xAI · proprietaryTB 83.3BC —OS —

83.3agentic / 100

Gemini 3.6 FlashGoogle · proprietaryTB —BC —OS 83

83agentic / 100

Holo3-35B-A3BH Company · open weightTB —BC —OS 82.56

82.6agentic / 100

Sakana Fugu-UltraSakana AI · proprietaryTB 82.1BC —OS —

82.1agentic / 100

Claude Sonnet 5Anthropic · proprietaryTB 80.4BC 84.7OS 81.2

81.9agentic / 100

GPT-5.5OpenAI · proprietaryTB 82BC 84.4OS 78.7

81.6agentic / 100

SWE-1.7Cognition · proprietaryTB 81.5BC —OS —

81.5agentic / 100

GLM-5.2Z.AI · open weightTB 81BC —OS —

81agentic / 100

Muse Spark 1.1Meta · proprietaryTB 80BC —OS 80.8

80.4agentic / 100

Claude Opus 4.8Anthropic · proprietaryTB 74.6BC 84.3OS 83.4

80.3agentic / 100

Sakana FuguSakana AI · proprietaryTB 80.2BC —OS —

80.2agentic / 100

Holo3-122B-A10BH Company · proprietaryTB —BC —OS 78.85

78.9agentic / 100

Ornith-1.0-397BDeepReinforce AI · open weightTB 77.5BC —OS —

77.5agentic / 100

Gemini 3.5 FlashGoogle · proprietaryTB 76.2BC —OS 78.4

77.2agentic / 100

GPT-5.4OpenAI · proprietaryTB 75.1BC 82.7OS 75

77.2agentic / 100

Agents-A1InternScience · open weightTB —BC 75.51OS —

75.5agentic / 100

Claude Opus 4.7 (Adaptive)Anthropic · proprietaryTB 69.4BC 79.3OS 78

75.1agentic / 100

DeepSeek V4 Pro 0813DeepSeek · proprietaryTB 67.9BC 83.4OS —

74.5agentic / 100

Kimi K2.6Moonshot AI · open weightTB 66.7BC 83.2OS 73.1

73.5agentic / 100

Claude Opus 4.6Anthropic · proprietaryTB 65.4BC 83.7OS 72.7

73agentic / 100

MiniMax M3MiniMax · open weightTB 66BC 83.52OS 70.06

72.3agentic / 100

Ling 3.0 FlashInclusionAI · open weightTB —BC 72.2OS —

72.2agentic / 100

Qwen3.7 PlusAlibaba · proprietaryTB 70.3BC —OS 73.3

71.7agentic / 100

GPT-5.3 CodexOpenAI · proprietaryTB 77.3BC —OS 64.7

71.4agentic / 100

Laguna S 2.1Poolside · open weightTB 70.2BC —OS —

70.2agentic / 100

Inkling-SmallThinking Machines Lab · open weightTB 64.7BC 77.4OS —

70.1agentic / 100

Qwen3.7 MaxAlibaba · proprietaryTB 69.7BC —OS —

69.7agentic / 100

InklingThinking Machines Lab · open weightTB 63.8BC 77.1OS —

69.4agentic / 100

Composer 2.5Cursor · proprietaryTB 69.3BC —OS —

69.3agentic / 100

MiMo-V2.5-ProXiaomi · proprietaryTB 68.4BC —OS —

68.4agentic / 100

Step 3.7 FlashStepFun · open weightTB 59.5BC 75.82OS —

66.4agentic / 100

Muse Glimmer 30BMeta · open weightTB —BC —OS 65.9

65.9agentic / 100

MiMo-V2.5Xiaomi · proprietaryTB 65.8BC —OS —

65.8agentic / 100

GPT-5.4 miniOpenAI · proprietaryTB 60BC —OS 72.1

65.7agentic / 100

GLM-5.1Z.AI · open weightTB 63.5BC 68OS —

65.4agentic / 100

Qwen 3.6 Max (preview)Alibaba · proprietaryTB 65.4BC —OS —

65.4agentic / 100

Claude Sonnet 4.6Anthropic · proprietaryTB 59.1BC —OS 72.1

65.2agentic / 100

Ornith-1.0-35BDeepReinforce AI · open weightTB 64.2BC —OS —

64.2agentic / 100

DeepSeek V4 Flash 0731DeepSeek · proprietaryTB 56.9BC 73.2OS —

63.8agentic / 100

Gemini 3.5 Flash-LiteGoogle · proprietaryTB 54BC —OS 74

63.4agentic / 100

Claude Opus 4.5Anthropic · proprietaryTB 59.3BC —OS 66.3

62.6agentic / 100

Composer 2Cursor · proprietaryTB 61.7BC —OS —

61.7agentic / 100

Qwen3.6 PlusAlibaba · proprietaryTB 61.6BC —OS —

61.6agentic / 100

Qwen3.6-27BAlibaba · open weightTB 59.3BC —OS —

59.3agentic / 100

Muse SparkMeta · proprietaryTB 59BC —OS —

59agentic / 100

MiniMax M2.7MiniMax · open weightTB 57BC —OS —

57agentic / 100

Qwen3.5 397BAlibaba · open weightTB 52.5BC 62OS —

56.5agentic / 100

Qwen3.5-122B-A10BAlibaba · open weightTB 49.4BC 63.8OS 58

56.4agentic / 100

GLM-5Z.AI · open weightTB 56.2BC —OS —

56.2agentic / 100

GPT-5.2OpenAI · proprietaryTB —BC 65.8OS 47.3

55.7agentic / 100

Claude Sonnet 4.5Anthropic · proprietaryTB 50BC —OS 61.4

55.4agentic / 100

Kimi K2.5Moonshot AI · open weightTB 50.8BC 60.6OS —

55agentic / 100

Kimi K2.5 (Reasoning)Moonshot AI · proprietaryTB 50.8BC 60.6OS —

55agentic / 100

Hy3 PreviewTencent · open weightTB 54.4BC —OS —

54.4agentic / 100

Qwen3.5-27BAlibaba · open weightTB 41.6BC 61OS 56.2

52agentic / 100

Qwen3.6-35B-A3BAlibaba · open weightTB 51.5BC —OS —

51.5agentic / 100

Nemotron 3 UltraNVIDIA · open weightTB 56.4BC 44.4OS —

51.3agentic / 100

Qwen3.5-35B-A3BAlibaba · open weightTB 40.5BC 61OS 54.5

51agentic / 100

Grok 4.20xAI · proprietaryTB 47.1BC —OS —

47.1agentic / 100

MAI-Thinking-1Microsoft · proprietaryTB 46BC —OS —

46agentic / 100

Laguna M.1Poolside · proprietaryTB 45.8BC —OS —

45.8agentic / 100

GLM-4.7Z.AI · open weightTB 41BC 52OS —

45.7agentic / 100

Ornith-1.0-9BDeepReinforce AI · open weightTB 43.1BC —OS —

43.1agentic / 100

GPT-5.4 nanoOpenAI · proprietaryTB 46.3BC —OS 39

42.9agentic / 100

Laguna XS.2Poolside · open weightTB 35.7BC —OS —

35.7agentic / 100

Nemotron 3.5 Lightning 30B A3B NVFP4NVIDIA · open weightTB 23.46BC 36.81OS —

29.1agentic / 100

Agentic score = weighted average of Terminal-Bench 2.0 (40%), OSWorld-Verified (35%), and BrowseComp (25%), normalized by available weights. This page intentionally stays on BenchLM's verified ranking lane and only includes exact-source rows. Display-only benchmarks (MCP Atlas, Toolathlon, etc.) are tracked but do not affect rankings.

Agentic Category Rankings

BenchLM category page with provisional and verified ranking modes

Coding Benchmarks

SWE-bench, LiveCodeBench, and more

Price vs Performance

Find the best value for agent workloads

Frequently Asked Questions

What are LLM agent benchmarks?

Agent benchmarks test whether AI models can go beyond answering questions and actually complete multi-step tasks: browsing the web, writing and running code in a terminal, calling external APIs via function calling, and operating desktop or mobile interfaces. They measure real-world usefulness for autonomous workflows.

What is function calling and why does it matter?

Function calling (or tool use) lets an LLM invoke external tools, APIs, or databases as part of its response. This is critical for building AI agents that can search the web, query databases, send emails, or control other software. Benchmarks like BFCL v4 and Toolathlon specifically measure how reliably models select the right function and pass correct arguments.

What is MCP (Model Context Protocol)?

MCP is an open standard for connecting LLMs to external tools and data sources. MCP Atlas and MCP-Tasks benchmark how well models work with MCP-backed integrations. Strong MCP performance means a model integrates well into tool-rich agent architectures.

Why does agentic carry the most weight in BenchLM scores?

Agentic carries 22% of BenchLM's overall score because the ability to use tools, browse, and complete multi-step tasks is the strongest differentiator between models in production use. A model that scores well on knowledge but cannot reliably call functions or navigate software has limited real-world utility for agent workflows.

Which models are best for building AI agents?

Currently, GPT-5.6 Sol by OpenAI leads BenchLM's verified agentic rankings with a score of 92. The best open-weight agent model is Qwen3.8 Max (86.1). Check the leaderboard above for the full verified ranking.

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.

Browse BenchLM

Tools & calculators

Tools