Artificial Analysis / Mercor

APEX-Agents-AA

Artificial Analysis' implementation of the APEX-Agents benchmark for long-horizon professional-services agent tasks.

In index
Agentic
Best in agentic
Source ↗

Models30

Top result47.1%Gemini 3.5 Flash

Top-3 spread5.8 pts

Year2026

Result and price

26 of 30 results

Results

30 results

#ModelResult

01Gemini 3.5 Flash47.1%

02Qwen3.8 Max Preview42.4%

03Kimi K341.3%

04GPT-5.6 Terra38.9%

05Step 5 Preview38.0%

06GPT-5.537.7%

07GPT-5.6 Luna35.8%

08GLM-5.233.7%

09GPT-5.433.3%

10Claude Opus 4.6 (Adaptive)33.0%

11Gemini 3.1 Pro32.0%

12Apodex 1.131.2%

12Apodex 1.1 Mini31.2%

14Kimi K2.628.5%

15GPT-5.4 mini28.2%

16Ling 3.0 Flash Fin27.4%

17GPT-5.4 nano24.9%

18DeepSeek V4 Pro 081324.3%

19Qwen3.7 Plus22.4%

20Grok 4.317.0%

21Qwen3.5 397B (Reasoning)15.3%

21Qwen3.5 397B15.3%

23Step 3.7 Flash14.8%

24GLM-514.5%

25Kimi K2.5 (Reasoning)11.5%

25Kimi K2.511.5%

27MiniMax M2.710.6%

28GPT-OSS 120B3.1%

29MiMo-V2.5-Pro2.4%

30GPT-OSS 20B0.7%