Name: 智能体模型厂商排行榜（2026-09-27）- 最佳 Agent 模型对比 - 厂商聚合排名
Creator: LM Arena
License: https://creativecommons.org/licenses/by/4.0/
Keywords: AI 智能体模型排行榜,智能体 AI 排行榜,最佳 Agent 模型对比,AI 模型排行,智能体评测,净提升

排名Rank ⇕	厂商Lab ⇕
› 🥇	AnthropicAnthropic Claude Fable 5.1 (Max) · Proprietary	12.32% ±3.22%

综合排名Overall Rank — 名次区间Rank Spread 1 ↔ 7 确认成功Confirmed Success ▲19.73%±4.79% 赞扬/抱怨Praise vs Complaint ▲23.93%±12.89% 可控性Steerability ▲5.03%±6.36% Bash 恢复Bash Recovery ▲12.63%±1.65% 工具幻觉Tool Hallucination ▼0.26%±0.03% 会话数Sessions 6,264
› 🥈	OpenAIOpenAI GPT 6 Astra (Max) · Proprietary	10.08% ±2.87%

综合排名Overall Rank — 名次区间Rank Spread 1 ↔ 10 确认成功Confirmed Success ▲13.06%±4.44% 赞扬/抱怨Praise vs Complaint ▲33.45%±11.14% 可控性Steerability ▲2.34%±6.53% Bash 恢复Bash Recovery ▲5.91%±1.46% 工具幻觉Tool Hallucination ▼0.35%±0.03% 会话数Sessions 6,874
› 🥉	SpaceXAISpaceXAI Grok 4.7 (xHigh) · Proprietary	4.38% ±1.68%

综合排名Overall Rank — 名次区间Rank Spread 7 ↔ 21 确认成功Confirmed Success ▲10.54%±3.06% 赞扬/抱怨Praise vs Complaint ▲5.57%±5.74% 可控性Steerability ▲0.31%±4.01% Bash 恢复Bash Recovery ▲5.76%±1.11% 工具幻觉Tool Hallucination ▼0.35%±0.03% 会话数Sessions 10,194
› 4	腾讯Tencent Hy4 preview · Apache 2.0	4.25% ±0.88%

综合排名Overall Rank — 名次区间Rank Spread 9 ↔ 20 确认成功Confirmed Success ▲7.65%±1.72% 赞扬/抱怨Praise vs Complaint ▲7.70%±3.17% 可控性Steerability ▲0.55%±1.84% Bash 恢复Bash Recovery ▲6.46%±0.50% 工具幻觉Tool Hallucination ▼0.01%±0.32% 会话数Sessions 53,124
› 5	月之暗面Moonshot Kimi K3 (Max) · Kimi K3 license	4.24% ±0.53%

综合排名Overall Rank — 名次区间Rank Spread 10 ↔ 19 确认成功Confirmed Success ▲9.50%±1.10% 赞扬/抱怨Praise vs Complaint ▲7.72%±1.82% 可控性Steerability ▲0.44%±1.13% Bash 恢复Bash Recovery ▲4.05%±0.49% 工具幻觉Tool Hallucination ▼0.35%±0.03% 会话数Sessions 129,565
› 6	GoogleGoogle Gemini 3.8 Flash (High) · Proprietary	5.85% ±1.81%

综合排名Overall Rank — 名次区间Rank Spread 3 ↔ 16 确认成功Confirmed Success ▲14.25%±3.02% 赞扬/抱怨Praise vs Complaint ▲10.58%±6.96% 可控性Steerability ▲2.32%±3.98% Bash 恢复Bash Recovery ▲1.95%±1.52% 工具幻觉Tool Hallucination ▼0.16%±0.14% 会话数Sessions 10,873
› 7	MetaMeta Muse Spark 1.3 (Max) · Proprietary	3.96% ±0.59%

综合排名Overall Rank — 名次区间Rank Spread 10 ↔ 20 确认成功Confirmed Success ▲8.32%±1.19% 赞扬/抱怨Praise vs Complaint ▲4.27%±2.04% 可控性Steerability ▲1.63%±1.30% Bash 恢复Bash Recovery ▲5.24%±0.57% 工具幻觉Tool Hallucination ▼0.35%±0.03% 会话数Sessions 62,677
› 8	DeepSeekDeepSeek Deepseek V4.1 Flash (Max) · MIT	3.96% ±0.57%

综合排名Overall Rank — 名次区间Rank Spread 10 ↔ 20 确认成功Confirmed Success ▲12.41%±1.11% 赞扬/抱怨Praise vs Complaint ▲2.95%±1.99% 可控性Steerability ▲3.29%±1.35% Bash 恢复Bash Recovery ▲7.50%±0.33% 工具幻觉Tool Hallucination ▼0.22%±0.04% 会话数Sessions 85,460
› 9	智谱 ZAIZ.ai GLM 5.2 (Max) · MIT · SiliconFlow	3.91% ±0.69%

综合排名Overall Rank — 名次区间Rank Spread 10 ↔ 21 确认成功Confirmed Success ▲3.91%±1.46% 赞扬/抱怨Praise vs Complaint ▲9.93%±2.40% 可控性Steerability ▲4.29%±1.32% Bash 恢复Bash Recovery ▲1.09%±0.69% 工具幻觉Tool Hallucination ▼0.35%±0.03% 会话数Sessions 83,775
› 10	阿里巴巴Alibaba Qwen3.8 Max · Proprietary	3.13% ±1.26%

综合排名Overall Rank — 名次区间Rank Spread 7 ↔ 26 确认成功Confirmed Success ▲6.44%±2.46% 赞扬/抱怨Praise vs Complaint ▲4.70%±4.57% 可控性Steerability ▲3.61%±2.63% Bash 恢复Bash Recovery ▲1.78%±1.60% 工具幻觉Tool Hallucination ▼0.85%±0.29% 会话数Sessions 19,589
› 11	小米Xiaomi Mimo V2.5 Pro · MIT	6.52% ±0.86%

综合排名Overall Rank — 名次区间Rank Spread 36 ↔ 41 确认成功Confirmed Success ▲10.98%±2.19% 赞扬/抱怨Praise vs Complaint ▲10.49%±2.49% 可控性Steerability ▲4.40%±1.69% Bash 恢复Bash Recovery ▲5.81%±1.51% 工具幻觉Tool Hallucination ▼0.92%±0.27% 会话数Sessions 44,443
› 12	MiniMaxMiniMax Minimax M3 · MiniMax Community License	6.77% ±0.84%

综合排名Overall Rank — 名次区间Rank Spread 36 ↔ 41 确认成功Confirmed Success ▲14.27%±2.28% 赞扬/抱怨Praise vs Complaint ▲13.18%±2.44% 可控性Steerability ▲6.96%±1.72% Bash 恢复Bash Recovery ▲0.63%±0.96% 工具幻觉Tool Hallucination ▼0.09%±0.19% 会话数Sessions 40,934
› 13	ThinkyThinky Inkling Small · Apache 2.0	9.59% ±1.35%

综合排名Overall Rank — 名次区间Rank Spread 41 ↔ 44 确认成功Confirmed Success ▲21.34%±4.01% 赞扬/抱怨Praise vs Complaint ▲20.37%±3.63% 可控性Steerability ▲11.62%±3.14% Bash 恢复Bash Recovery ▲5.79%±1.23% 工具幻觉Tool Hallucination ▼0.42%±0.33% 会话数Sessions 13,067
› 14	MistralMistral Mistral Medium 3.5 · Modified MIT	11.65% ±1.46%

综合排名Overall Rank — 名次区间Rank Spread 42 ↔ 44 确认成功Confirmed Success ▲15.84%±3.68% 赞扬/抱怨Praise vs Complaint ▲15.12%±4.02% 可控性Steerability ▲6.88%±3.25% Bash 恢复Bash Recovery ▲18.11%±2.71% 工具幻觉Tool Hallucination ▼2.30%±0.73% 会话数Sessions 10,822
› 15	UpstageUpstage Solar Pro 4 · Proprietary	16.14% ±1.87%

综合排名Overall Rank — 名次区间Rank Spread 45 ↔ 45 确认成功Confirmed Success ▲23.06%±4.88% 赞扬/抱怨Praise vs Complaint ▲20.19%±4.88% 可控性Steerability ▲9.82%±3.69% Bash 恢复Bash Recovery ▲25.71%±4.17% 工具幻觉Tool Hallucination ▼1.90%±0.74% 会话数Sessions 9,284

没有找到相关模型No matching models found

AI 智能体模型排行榜Agent Models Leaderboard 🏆 综合榜单