Interactive Agent Task on TRIP-Bench (Hard LIT) benchmark leaderboard · SOTA2 Research

Interactive Agent Task on TRIP-Bench (Hard LIT)

4,400Loose SuccessGPT-5.2

Evaluation Results

Method	Links
GPT-5.22026.02		4,400	1,400
DeepSeek-V3.22026.02		3,600	200
Claude-Sonnet-4.52026.02		2,800	0
DeepSeek-V3.22026.02		1,600	200
Gemini-3-Pro2026.02		1,600	0
Gemini-3-Pro2026.02		1,200	0
GPT-5.22026.02		1,000	0
Claude-Sonnet-4.52026.02		1,000	0
GLM-4.72026.02		1,000	0
Gemini-3-Flash2026.02		1,000	0
Gemini-3-Flash2026.02		800	0
GLM-4.72026.02		600	0
Gemini-3-Flash2026.02		26	0
DeepSeek-V3.22026.02		26	0
Claude-Sonnet-4.52026.02		22	0
GPT-5.22026.02		20	0
Gemini-3-Pro2026.02		18	0
Gemini-3-Flash2026.02		16	0
Gemini-3-Pro2026.02		12	0
Claude-Sonnet-4.52026.02		10	0
GPT-5.22026.02		8	0
Qwen3-235B-A22B-Instruct-25072026.02		2	0
DeepSeek-V3.22026.02		2	0
Kimi-K2-0905-Preview2026.02		0	0
Qwen3-235B-A22B-Instruct-25072026.02		0	0
Qwen3-235B-A22B-Thinking-25072026.02		0	0
Kimi-K2-Thinking2026.02		0	0
Kimi-K2-0905-Preview2026.02		0	0
GLM-4.72026.02		0	0
Qwen3-235B-A22B-Thinking-25072026.02		0	0
Kimi-K2-Thinking2026.02		0	0
GLM-4.72026.02		0	0