Claw-Eval leaderboard

A transparent real-world autonomous-agent benchmark with 300 human-verified tasks, 2,159 rubric items, and Pass^3 scoring across general, multi-turn, and native multimodal agent tasks.

Models ranked by their published Claw-Eval value

Rank	Model	Provider	End-to-end autonomous-agent evaluation with Pass^3 scoring
1	OAOrnith-1.5-397B	Ornith AI	81.4
2	K-EXAONE 2.0	LG AI Research	77.7
3	DAOrnith-1.0-397B	DeepReinforce AI	77.1
4	MiniMax M3	MiniMax	74.5
5	DSdots3-note Preview	Dots Studio	73.4
6	OAOrnith-1.5-35B-A3B	Ornith AI	72.5
7	Qwen3.6-27B	Alibaba	72.4
8	Claude Opus 4.6	Anthropic	70.4
9	DAOrnith-1.0-35B	DeepReinforce AI	69.8
10	Qwen3.6-35B-A3B	Alibaba	68.7
11	Claude Sonnet 4.6	Anthropic	67.8
12	Step 3.7 Flash	StepFun	67.1
13	OAOrnith-1.5-9B	Ornith AI	66.5
14	Qwen3.7 Max	Alibaba	65.2
15	LLaDA2.2-flash	InclusionAI	64.2
16	MiMo-V2.5-Pro	Xiaomi	63.8
16	Muse Spark	Meta	63.8
18	DAOrnith-1.0-9B	DeepReinforce AI	63.1
19	LFM2.5-2.6B	LiquidAI	62.9
20	Qwen3.7 Plus	Alibaba	62.7
21	GLM-5.1	Z.AI	62.3
21	Kimi K2.6	Moonshot AI	62.3
21	MiMo-V2.5	Xiaomi	62.3
24	GPT-5.4	OpenAI	60.3
25	Claude Opus 4.5	Anthropic	59.6
26	Qwen3.6 Plus	Alibaba	58.8
27	Gemini 3.1 Pro	Google	57.8
27	MiMo-V2-Pro	Xiaomi	57.8
29	GLM-5	Z.AI	57.7
30	LLaDA2.2-mini	InclusionAI	57.2
31	Qwen3.5 397B	Alibaba	56.8
32	GLM-5-Turbo	Z.AI	55.8
33	GLM-5V-Turbo	Z.AI	53.8
34	Kimi K2.5	Moonshot AI	52.3
35	Gemini 3 Flash	Google	49.2
36	MiniMax M2.7	MiniMax	48.7
37	MiMo-V2-Omni	Xiaomi	45.2
38	DeepSeek V3.2	DeepSeek	40.2
39	Nemotron 3 Super 100B	NVIDIA	5.5

Evidence key: Observed

Rows are ordered by the value Claw-Eval: Towards Trustworthy Evaluation of Autonomous Agents published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.