Benchmark Explorer

Explore how models perform on various benchmarks

Benchmarks

🔍

τ²-Bench Telecom

A dual-control conversational AI benchmark simulating technical support scenarios where both agent and user must coordinate actions to resolve telecom service issues. The benchmark introduces a new paradigm for evaluating conversational AI by simulating both the agent and user to actively modify a shared world state.
Source:

Model Performance

#1
Grok 4.20 (Reasoning)
97.0%
#2
GLM-5
97.0%
#3
Gemini 3.1 Pro Preview
96.0%
#4
MiniMax M2.5
95.0%
#5
GLM-4.7
94.0%
#6
Claude Opus 4.6 (Thinking)
92.0%
#7
GPT-5.2 Codex
92.0%
#8
GPT-5.3 Codex
91.0%
#9
DeepSeek V3.2 (Thinking)
91.0%
#10
Claude Opus 4.5 (Thinking)
90.0%
#11
Grok 3 Mini
90.0%
#12
Claude Opus 4.5
86.0%
#13
GPT-5 (Thinking)
85.0%
#14
Claude Opus 4.6
85.0%
#15
MiniMax M2.7
85.0%
#16
MiniMax M2.1
85.0%
#17
MiMo V2 Flash
84.0%
#18
Qwen 3 Max (Thinking)
84.0%
#19
GPT-5.1 Codex Max
83.0%
#20
Kimi K2.5
81.0%
#21
OpenAI o3 (Medium Effort)
81.0%
#22
OpenAI o3 (High Effort)
81.0%
#23
Claude Sonnet 4.6
80.0%
#24
DeepSeek V3.2
79.0%
#25
Claude Sonnet 4.5 (Thinking)
78.0%
#26
Grok 4
75.0%
#27
Grok 4 (Thinking)
75.0%
#28
Claude 4.0 Opus (Thinking)
73.0%
#29
Claude 4.1 Opus (Thinking)
71.0%
#30
Claude Sonnet 4.5
71.0%
#31
Grok 4.20
70.0%
#32
GPT-5 Mini (Thinking)
68.0%
#33
GPT OSS 120B
66.0%
#34
Claude 4.0 Sonnet (Thinking)
65.0%
#35
OpenAI o1
63.0%
#36
Kimi K2
61.0%
#37
OpenAI o4 Mini (High Effort)
56.0%
#38
OpenAI o4 Mini (Medium Effort)
56.0%
#39
Claude 3.7 Sonnet (Thinking)
55.0%
#40
Claude Haiku 4.5 (Thinking)
55.0%
#41
Gemini 2.5 Pro (Thinking)
54.0%
#42
Qwen 3 (Thinking)
53.0%
#43
Claude 4.0 Sonnet
52.0%
#44
Claude 3.7 Sonnet
50.0%
#45
Grok 3
49.0%
#46
DeepSeek V3 (Mar 2025)
47.0%
#47
GPT-4.1
47.0%
#48
GPT-5.2
47.0%
#49
Gemini 3.0 Flash
43.0%
#50
Mistral Medium 3.1
41.0%
#51
DeepSeek R1
37.0%
#52
GPT-5 Nano (Thinking)
37.0%
#53
GPT-5.4
35.0%
#54
GPT-5.4 Nano
35.0%
#55
Qwen 3
33.0%
#56
Claude Haiku 4.5
33.0%
#57
Qwen 3 Max Preview
33.0%
#58
GPT-5 Mini
32.0%
#59
Gemini 2.5 Flash (Thinking)
32.0%
#60
OpenAI o3 Mini (High Effort)
31.0%
#61
Gemini 3.1 Flash Lite Preview
31.0%
#62
Mistral Large 2
31.0%
#63
Gemini 2.0 Flash
30.0%
#64
OpenAI o3 Mini (Medium Effort)
29.0%
#65
Llama 3.3 70B
27.0%
#66
GPT-5 Nano
26.0%
#67
Claude 3.5 Haiku
25.0%
#68
Devstral 2
25.0%
#69
Mistral Large 3
25.0%
#70
DeepSeek V3
23.0%
#71
Magistral Medium 3.1
23.0%
#72
GPT-5.4 Mini
23.0%
#73
Devstral Small 2
23.0%
#74
Llama 3.1 405B
19.0%
#75
Llama 4 Maverick 17B
18.0%
#76
Cohere Command A
15.0%
#77
Gemini 2.5 Flash
15.0%
#78
Phi-4
0.0%