Inference
EN

Agents' Last Exam

An agent benchmark reported in DeepSeek's V4 Flash 0731 launch comparison.

ModelsModels with a result.21
Top resultBest result on this benchmark.60.0%GPT-6.1 Terra
Top-3 spreadPoints from first to third.3.6 pts
YearYear of release.—

Result and price

19 of 21 results

Results

21 results
#ModelResultThe score from the source.
01GPT-6.1 Terra60.0%
02GPT-6 Astra59.3%
03GPT-6 Sol56.4%
04Qwen3.8 MaxAbove fitFar above this model's index.52.4%
05Qwen3.8-Flash-NextAbove fitFar above this model's index.51.2%
06Kimi K3.1 Pro43.2%
07Qwen3.8-27BAbove fitFar above this model's index.42.9%
08Hy441.0%
09Gemini 4 Argon39.5%
10LongCat-2.538.5%
11DeepSeek V4.1 Flash 101735.4%
12MiMo-V2.6-Pro31.6%
13MiniMax M3.1 Flash29.6%
14Step 5 Preview29.5%
15GLM-5.328.5%
16MiMo-V2.6-Flash27.6%
17Gemini 3.7 Flash26.3%
17GLM-5.3-Flash26.3%
19DeepSeek V4 Pro 081325.7%
20DeepSeek V4 Flash 073125.2%
21Hy4 preview22.8%