τ²-Bench Telecom

Artificial Analysis · 50 models

Rank	Model	Score
1	GLM-5.2 (Max)	0.9912
1	JT-35B-Flash	0.9912
3	GLM-4.7-Flash (Reasoning)	0.9883
4	Claude Fable 5 (Max, Opus 4.8 Fallback)	0.9854
4	GLM 5V Turbo (Reasoning)	0.9854
4	GLM-5-Turbo	0.9854
4	Step 3.7 Flash	0.9854
8	GLM-5 (Reasoning)	0.9825
9	GLM-5.1 (Reasoning)	0.9766
9	Grok 4.3 (High)	0.9766
9	Qwen3.6 Plus	0.9766
12	GLM-5 (Non-reasoning)	0.9737
13	GLM-5.1 (Non-reasoning)	0.9708
14	Grok 4.20 0309 (Reasoning)	0.9649
15	DeepSeek V4 Pro 0424 (Max)	0.962
16	GLM-4.7 (Reasoning)	0.9591
16	Kimi K2.5 (Reasoning)	0.9591
16	Kimi K2.6 (Reasoning)	0.9591
16	Qwen3.6 Max Preview	0.9591
20	DeepSeek V4 Flash 0420 (High)	0.9561
20	Gemini 3.1 Pro Preview	0.9561
20	Gemini 3.5 Flash (Medium)	0.9561
20	Qwen3.5 397B A17B (Reasoning)	0.9561
24	Gemini 3.5 Flash (High)	0.9532
24	MiniMax-M2.5	0.9532
24	Qwen3.6 35B A3B (Reasoning)	0.9532
27	DeepSeek V4 Flash 0420 (Max)	0.9503
27	MiMo-V2-Flash (Reasoning)	0.9503
27	MiMo-V2-Pro	0.9503
30	Qwen3.7 Max	0.9474
31	Claude Opus 4.8 (Max)	0.9444
31	DeepSeek V4 Flash 0420 (Non-reasoning)	0.9444
33	Step 3.5 Flash	0.9444
34	DeepSeek V4 Pro 0424 (High)	0.9415
34	GLM-4.7 (Non-reasoning)	0.9415
34	MiMo-V2.5-Pro (Reasoning)	0.9415
34	Mistral Medium 3.5	0.9415
34	Qwen3.6 27B (Reasoning)	0.9415
39	GPT-5.5 (Xhigh)	0.9386
39	Kimi K2.6 (Non-reasoning)	0.9386
39	Qwen3.5 27B (Reasoning)	0.9386
42	Qwen3.5 122B A10B (Reasoning)	0.9357
42	Qwen3.6 27B (Non-reasoning)	0.9357
44	Grok 4.1 Fast (Reasoning)	0.9327
44	MiMo-V2-Flash (Feb 2026)	0.9327
44	Tri-21B-think Preview	0.9327
47	GPT-5.5 (High)	0.9298
47	Grok 4.20 0309 v2 (Reasoning)	0.9298
47	JT-MINI	0.9298
47	Kimi K2 Thinking	0.9298