Autonomous Agent Spatio-Temporal Reasoning on STT-Arena benchmark leaderboard · SOTA2 Research

Autonomous Agent Spatio-Temporal Reasoning on STT-Arena

41.78Pass@1 (Easy)Claude-4.6-Opus

Evaluation Results

Method	Links
Claude-4.6-Opus2026.05		41.78	31.34	23.73	52.22	35.39
GPT-5.42026.05		39.91	29.85	24.29	52.22	34.51
Gemini-3.1-Pro2026.05		37.56	30.35	21.47	52.22	33.19
Deepseek-V3.22026.05		36.62	27.86	15.82	63.33	32.16
Qwen-3.6-Plus2026.05		35.21	24.38	23.16	54.44	31.42
Claude-4.6-Sonnet2026.05		35.21	29.85	19.77	58.89	32.74
GLM-5.12026.05		35.21	23.38	15.25	41.11	27.31
Qwen-3.5-397B-A17B2026.05		33.8	21.89	17.51	54.45	28.78
Deepseek-V4-Pro2026.05		32.86	20.4	15.25	45.56	26.29
GLM-52026.05		30.05	21.39	10.73	43.33	24.23
STT-Agent-4B (SFT+RL)2026.05		29.11	19.9	14.12	64.44	27.17
Gemini-2.5-flash2026.05		28.17	13.43	12.43	31.11	20.11
Qwen-3.5-9B2026.05		27.23	12.93	11.3	32.22	19.53
MiniMax-M2.52026.05		26.76	15.92	12.43	42.22	21.88
STT-Agent-4B (SFT)2026.05		26.76	17.41	13.56	61.11	25.11
MiniMax-M2.72026.05		26.29	15.92	13.56	36.67	21.29
Kimi-K2.52026.05		25.35	13.43	14.12	37.78	19.18
Llama-3.3-70B2026.05		22.07	11.44	11.3	30	17.18
Qwen-3.5-35B-A3B2026.05		19.72	13.43	7.34	17.78	12.48
GPT-5.4-mini2026.05		19.72	13.43	7.34	6.67	12.92
Qwen-3-4B (baseline)2026.05		18.31	9.46	2.82	10	10.57
Qwen-3-8B2026.05		16.9	13.43	11.3	20	14.83
Llama-3.1-8B2026.05		7.98	3.98	2.82	5.56	5.14