Arena — Agent

Name: Arena — Agent
Creator: Arena

按 Arena 發布的原樣呈現，指標和模型都以它為準。

發布方: Arena
授權條款: 未註明
更新頻率: 持續更新
資料截至: 2026-10-08
本站抓取時間: 2026-10-09
條目數: 52

按 Overall 排名前 12 的模型

Claude Opus 5.5 0.143
GPT 6 Astra 0.131
Claude Fable 5.1 0.127
Claude Sonnet 5.5 0.120
GPT 6.1 Sol 0.117
GPT 6 Sol 0.099
Gemini 4 Argon 0.093
Claude Fable 5 0.089
Claude Opus 5 0.081
Claude Opus 4.8 0.071
GPT 5.6 Sol 0.062
Claude Sonnet 5 0.049

排序依據只看已識別的模型

52 個有分數

#	模型
1	Claude Opus 5.5agent · highAnthropic	0.1430.122–0.165	0.139	0.337	0.100	0.136	0.005
2	GPT 6 Astraagent · maxOpenAI	0.1310.107–0.155	0.121	0.385	0.088	0.055	0.004
3	Claude Fable 5.1agent · maxAnthropic	0.1270.106–0.147	0.162	0.262	0.088	0.116	0.005
4	Claude Sonnet 5.5agent · maxAnthropic	0.1200.086–0.153	0.158	0.234	0.051	0.150	0.004
5	GPT 6.1 Solagent · maxOpenAI	0.1170.087–0.148	0.147	0.303	0.085	0.045	0.005
6	GPT 6 Solagent · maxOpenAI	0.0990.075–0.123	0.070	0.236	0.119	0.070	0.001
7	Gemini 4 Argonagent · highGoogle	0.0930.069–0.116	0.156	0.406	0.116	-0.214	-0.002
8	Claude Fable 5agent · highAnthropic	0.0890.077–0.102	0.058	0.193	0.095	0.096	0.005
9	Claude Opus 5agent · maxAnthropic	0.0810.066–0.096	0.104	0.119	0.042	0.136	0.005
10	Claude Opus 5agent · highAnthropic	0.0800.066–0.093	0.084	0.124	0.058	0.126	0.005
11	Claude Opus 4.8agent · highAnthropic	0.0710.058–0.085	0.047	0.151	0.084	0.072	0.002
12	GPT 5.6 Solagent · xhighOpenAI	0.0620.049–0.075	0.053	0.183	0.065	0.007	0.003
13	Claude Sonnet 5agent · highAnthropic	0.0490.032–0.066	0.022	0.101	0.046	0.074	0.003
14	GPT 5.5agent · xhighOpenAI	0.0440.033–0.055	-0.009	0.088	0.041	0.095	0.004
15	Muse Spark 1.3agent · maxMeta	0.0400.034–0.047	0.053	0.036	0.048	0.061	0.004
16	DeepSeek V4.1 Flashagent · maxDeepSeek	0.0380.034–0.042	0.074	0.024	0.009	0.079	0.004
17	Kimi K3agent · maxKimi	0.0380.032–0.044	0.065	0.064	0.011	0.044	0.005
18	Hy4agent · previewTencent	0.0360.029–0.043	0.060	0.052	0.003	0.065	0.001
19	MiMo V2.6 ProagentXiaomi	0.0350.016–0.054	0.080	0.082	-0.006	0.023	-0.004
20	GLM 5.2agent · max · SiliconFlowZ AI	0.0320.025–0.039	0.041	0.089	0.017	0.007	0.005
21	Grok 4.7agent · xhighSpaceXAI	0.0280.011–0.045	0.039	0.051	-0.024	0.072	0.004
22	Gemini 3.8 Flashagent · highGoogle	0.0280.018–0.037	0.053	0.063	0.025	-0.007	0.003
23	Qwen 3.8 MaxunknownAlibaba	0.0230.018–0.029	0.051	0.029	0.018	0.028	-0.009
24	GLM 5.3agent · maxZ AI	0.0210.014–0.029	0.048	0.051	0.004	-0.001	0.004
25	GPT 6 Lunaagent · maxOpenAI	0.0190.013–0.026	0.009	0.003	0.031	0.052	0.001
26	DeepSeek V4 Prounknown · 0813 · highDeepSeek	0.013-0.003–0.030	-0.020	0.019	0.021	0.044	0.003
27	Grok 4.5agentSpaceXAI	0.0120.002–0.022	0.010	-0.035	0.029	0.051	0.005
28	GPT 5.4arena-harness · highOpenAI	0.0120.002–0.021	-0.028	0.008	0.065	0.012	0.002
29	Grok 4.6agent · xhighSpaceXAI	0.0110.001–0.021	-0.008	-0.007	0.007	0.057	0.004
30	GPT 5.5agentOpenAI	0.009-0.001–0.019	-0.051	0.037	0.017	0.041	0.003
31	GPT 5.6 Terraagent · xhighOpenAI	0.009-0.003–0.021	-0.032	0.036	0.028	0.009	0.002
32	MiMo V2.6 FlashagentXiaomi	0.004-0.011–0.018	0.054	0.018	-0.004	-0.011	-0.039
33	Step 5agent · previewStepFun	0.000-0.016–0.016	0.024	-0.030	-0.001	0.002	0.005
34	Qwen 3.8 Flash NextagentAlibaba	-0.007-0.013–-0.001	0.038	-0.034	-0.007	-0.026	-0.007
35	GLM 5.3 FlashagentZ AI	-0.010-0.015–-0.004	0.035	-0.018	-0.012	-0.057	0.004
36	GPT 5.6 Lunaagent · xhighOpenAI	-0.015-0.022–-0.008	-0.042	-0.039	-0.000	0.005	0.003
37	Qwen 3.8agent · 27BAlibaba	-0.020-0.026–-0.014	0.008	-0.039	-0.008	-0.059	-0.003
38	Gemini 3.7 Flashagent · highGoogle	-0.021-0.029–-0.013	-0.002	-0.038	-0.013	-0.054	0.002
39	Muse Spark 1.2agent · xhighMeta	-0.037-0.044–-0.029	-0.067	-0.122	-0.051	0.058	-0.000
40	Muse Spark 1.1agentMeta	-0.050-0.055–-0.045	-0.042	-0.120	-0.063	-0.023	-0.001
41	Hunyuan Hy3agentTencent	-0.051-0.063–-0.038	-0.117	-0.059	0.021	-0.072	-0.027
42	Qwen 3.7 MaxagentAlibaba	-0.055-0.065–-0.045	-0.100	-0.098	-0.061	-0.016	0.000
43	Mistral Large 4agentMistral	-0.066-0.095–-0.036	0.017	-0.044	-0.133	0.018	-0.186
44	MiniMax M3first-partyMiniMax	-0.070-0.079–-0.061	-0.147	-0.140	-0.067	0.006	-0.000
45	Qwen 3.7 PlusagentAlibaba	-0.072-0.087–-0.058	-0.091	-0.112	-0.086	-0.063	-0.009
46	MiMo V2.5 ProagentXiaomi	-0.078-0.087–-0.069	-0.090	-0.128	-0.087	-0.075	-0.009
47	Gemini 3.6 Flashagent · highGoogle	-0.078-0.089–-0.067	-0.058	-0.089	-0.065	-0.182	0.005
48	Gemini 3.1 Proagent · previewGoogle	-0.078-0.088–-0.068	-0.101	-0.013	-0.040	-0.233	-0.003
49	Inkling SmallunknownThinking Machines	-0.103-0.118–-0.089	-0.202	-0.225	-0.132	0.054	-0.011
50	InklingagentThinking Machines	-0.115-0.126–-0.104	-0.217	-0.233	-0.122	0.001	-0.004
51	Mistral Medium 3.5agentMistral	-0.126-0.141–-0.111	-0.144	-0.196	-0.062	-0.204	-0.024
52	Solar Pro 4agentUpstage	-0.170-0.188–-0.151	-0.230	-0.245	-0.100	-0.256	-0.018

每一列測的是什麼

Overall: 越高越好 · 代理式工作
Task outcome: 越高越好
Praise and complaint: 越高越好
Steering burden: 越高越好
Recovery after a failed command: 越高越好
Tool hallucination: 越高越好