Arena — Agent

As Arena publishes it — their metrics, their models. Follow a model name to its page upstream.

Published by

Arena

Licence

not stated

Updates

continuous

Data as of

2026-08-19

Fetched

2026-08-21

Entries

51 · 6 unmatched

Filter modelsOrder byOverallTask outcomePraise and complaintSteerabilityRecovery after a failed commandTool hallucinationMapped models only

51 scored

#	Model	Overall ↑	Task outcome ↑	Praise and complaint ↑	Steerability ↑	Recovery after a failed command ↑	Tool hallucination ↑
1	Claude Opus 5unknown · highanthropic	0.1250.109–0.140	0.154	0.205	0.112	0.142	0.010
2	Claude Opus 5unknown · maxanthropic	0.1200.102–0.138	0.185	0.191	0.067	0.146	0.011
3	Claude Fable 5 (High)unknown · highUnmatched	0.1160.099–0.133	0.124	0.214	0.090	0.139	0.011
4	Kimi K3unknown · maxkimi	0.1040.098–0.110	0.180	0.183	0.060	0.087	0.011
5	GPT-5.6 Solunknown · xhighopenai	0.0970.083–0.111	0.100	0.222	0.058	0.096	0.011
6	Claude Opus 4.8unknown · highanthropic	0.0950.080–0.111	0.083	0.217	0.090	0.092	-0.004
7	GPT-5.5unknown · xhighopenai	0.0850.076–0.094	0.044	0.135	0.089	0.147	0.011
8	Claude Opus 4.7unknown · highanthropic	0.0810.069–0.094	0.056	0.124	0.084	0.132	0.010
9	GPT-5.5unknown · highopenai	0.0770.069–0.086	0.041	0.113	0.091	0.131	0.011
10	Claude Opus 4.7unknownanthropic	0.0740.062–0.086	0.050	0.115	0.093	0.101	0.011
11	Claude Sonnet 5unknown · highanthropic	0.0660.044–0.088	0.015	0.148	0.046	0.111	0.010
12	Claude Opus 4.6unknownanthropic	0.0660.054–0.078	0.049	0.085	0.075	0.110	0.011
13	GPT-5.5unknownopenai	0.0640.056–0.072	0.033	0.078	0.079	0.117	0.011
14	DeepSeek V4 Pro (High) (0813)unknownUnmatched	0.0630.050–0.075	0.131	0.038	0.024	0.109	0.011
15	Qwen3.8 Maxunknownalibaba	0.0620.048–0.076	0.116	0.068	0.040	0.084	0.002
16	Grok 4.5unknownxai	0.0620.050–0.073	0.055	0.056	0.072	0.115	0.011
17	GLM-5.2unknown · max · SiliconFlowzai	0.0580.050–0.067	0.072	0.094	0.052	0.061	0.011
18	GPT-5.4unknown · highopenai	0.0500.042–0.058	0.046	0.031	0.066	0.095	0.011
19	GPT-5.6 Lunaunknown · xhighopenai	0.0400.022–0.059	-0.019	0.078	0.017	0.114	0.011
20	Deepseek V4 Flash (High) (20260731)unknownUnmatched	0.0400.032–0.048	0.081	0.021	0.035	0.051	0.011
21	Gemini 3.7 Flashunknown · highgoogle	0.0330.023–0.043	0.100	-0.000	0.027	0.029	0.011
22	GPT-5.6 Terraunknown · xhighopenai	0.0320.020–0.044	-0.018	0.018	0.050	0.099	0.011
23	Claude Sonnet 4.6unknownanthropic	0.0290.017–0.041	-0.007	0.010	0.018	0.112	0.010
24	Claude Opus 4.8unknownanthropic	0.0250.003–0.048	0.078	0.137	0.081	0.104	-0.273
25	Muse Spark 1.2unknown · xhighmeta	0.0210.009–0.032	0.060	-0.057	-0.025	0.114	0.011
26	Muse Spark 1.1unknownmeta	0.0090.003–0.015	0.066	-0.056	-0.034	0.059	0.011
27	Kimi K2.7 Codeunknownkimi	0.004-0.016–0.025	0.037	0.015	-0.026	-0.016	0.011
28	DeepSeek V4 Prounknowndeepseek	0.001-0.008–0.009	-0.027	-0.024	0.002	0.048	0.003
29	GLM-5.1unknown · SiliconFlowzai	-0.000-0.008–0.007	0.009	-0.003	0.010	-0.011	-0.006
30	Gemini 3.5 Flashunknown · highgoogle	-0.007-0.013–-0.001	0.001	-0.006	-0.005	-0.025	0.001
31	Qwen3.7 Maxunknownalibaba	-0.008-0.016–0.001	-0.027	-0.062	-0.004	0.048	0.006
32	Gemini 3.1 Pro Previewunknowngoogle	-0.009-0.016–-0.002	0.008	0.029	0.032	-0.122	0.009
33	Kimi K2.6unknownkimi	-0.011-0.033–0.011	-0.015	0.015	-0.006	-0.060	0.011
34	MiMo-V2.5-Prounknownxiaomi	-0.023-0.031–-0.014	-0.033	-0.078	-0.017	0.015	0.001
35	Hy3unknowntencent	-0.024-0.036–-0.011	-0.019	-0.021	-0.090	0.023	-0.013
36	Qwen3.7 Plusunknownalibaba	-0.025-0.038–-0.012	-0.030	-0.103	-0.056	0.061	0.001
37	Gemini 3.6 Flashunknown · highgoogle	-0.030-0.043–-0.018	-0.035	-0.051	-0.034	-0.043	0.011
38	MiniMax-M3unknownminimax	-0.031-0.040–-0.023	-0.066	-0.093	-0.062	0.058	0.006
39	Gemini 3.5 Flashunknown · mediumgoogle	-0.039-0.053–-0.025	-0.087	-0.064	-0.032	-0.015	0.005
40	Inkling Smallunknownthinking-machines	-0.068-0.086–-0.051	-0.176	-0.148	-0.128	0.107	0.004
41	Inklingunknownthinking-machines	-0.072-0.081–-0.063	-0.122	-0.179	-0.137	0.072	0.006
42	Mistral Medium 3.5unknownmistral	-0.076-0.095–-0.057	-0.115	-0.113	-0.112	-0.017	-0.023
43	Grok 4.3unknown · highxai	-0.090-0.099–-0.082	-0.105	-0.139	-0.073	-0.143	0.009
44	Gemini 3 FlashunknownUnmatched	-0.091-0.099–-0.083	-0.077	-0.107	-0.038	-0.229	-0.006
45	Grok Build 0.1unknownUnmatched	-0.095-0.104–-0.086	-0.056	-0.109	-0.092	-0.224	0.007
46	Solar Pro 4unknownupstage	-0.108-0.132–-0.084	-0.071	-0.172	-0.160	-0.128	-0.010
47	Gemini 3.5 Flash-Liteunknowngoogle	-0.108-0.121–-0.096	-0.128	-0.141	-0.112	-0.159	-0.002
48	MiniMax-M2.7unknownminimax	-0.119-0.130–-0.108	-0.123	-0.164	-0.143	-0.173	0.010
49	Nemotron 3 UltraunknownUnmatched	-0.154-0.177–-0.130	-0.153	-0.155	-0.205	-0.259	0.003
50	Grok 4.3unknownxai	-0.165-0.177–-0.152	-0.113	-0.160	-0.057	-0.502	0.010
51	Gemma 4 31Bunknowngoogle	-0.198-0.224–-0.172	-0.013	-0.035	-0.084	-0.541	-0.318

What the columns measure

Overall

Higher is better · Agentic work

Task outcome

Higher is better

Praise and complaint

Higher is better

Steerability

Higher is better

Recovery after a failed command

Higher is better

Tool hallucination

Higher is better