Skip to main content
benchlm.ai

Tau2-Telecom

A telecom-oriented tool benchmark that measures structured tool use in domain workflows.

Benchmark score on Tau2-Telecom — June 13, 2026

BenchLM mirrors the published score view for Tau2-Telecom. Step 3.7 Flash leads the public snapshot at 98.5% , followed by GLM-5V-Turbo (98.5%) and GLM-5-Turbo (98.5%). BenchLM does not use these results to rank models overall.

1Open

Step 3.7 Flash

StepFun

98.5%
Overall —Context 256K
2Closed

GLM-5V-Turbo

Z.AI

98.5%
Overall —Context 200K
3Closed

GLM-5-Turbo

Z.AI

98.5%
Overall —Context 200K
123 modelsAgenticCurrentDisplay onlyUpdated June 13, 2026

The published Tau2-Telecom snapshot is tightly clustered at the top: Step 3.7 Flash sits at 98.5%, while the third row is only 0.0 points behind. The broader top-10 spread is 2.6 points, so many of the published scores sit in a relatively narrow band.

123 models have been evaluated on Tau2-Telecom. The benchmark falls in the Agentic category. This category carries a 22% weight in BenchLM.ai's overall scoring system. Tau2-Telecom is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About Tau2-Telecom

Year

2026

Tasks

Telecom tool workflows

Format

Domain-specific tool evaluation

Difficulty

Professional workflow

OpenAI reports tau2-bench as a domain-specific tool benchmark for telecom tasks, useful for measuring API-call reliability under constraints.

Introducing GPT-5.4 mini and nano

BenchLM freshness & provenance

Version

τ²-Bench 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Benchmark score table (123 models)

1
Step 3.7 Flash
98.5%
2
GLM-5V-Turbo
98.5%
3
GLM-5-Turbo
98.5%
4
GLM-5
98.2%
5
GLM-5.1
97.7%
6
Qwen3.6 Plus
97.7%
7
Grok 4.3
97.7%
8
DeepSeek V4 Pro (Max)
96.2%
9
Kimi K2.6
95.9%
10
Kimi K2.5 (Reasoning)
95.9%
11
GLM-4.7
95.9%
12
Kimi K2.5
95.9%
13
Qwen 3.6 Max (preview)
95.9%
14
Gemini 3.1 Pro
95.6%
15
Qwen3.5 397B (Reasoning)
95.6%
16
DeepSeek V4 Flash (High)
95.6%
17
Gemini 3.5 Flash
95.3%
18
Qwen3.6-35B-A3B
95.3%
19
DeepSeek V4 Flash (Max)
95%
20
MiMo-V2-Pro
95%
21
Qwen3.7 Max
94.7%
22
Claude Opus 4.8
94.4%
23
DeepSeek V4 Pro (High)
94.2%
24
Qwen3.6-27B
94.2%
25
MiMo-V2.5-Pro
94.2%
26
Mistral Medium 3.5 128B
94.2%
27
GPT-5.5
93.9%
28
Qwen3.5-27B
93.9%
29
Qwen3.5-122B-A10B
93.6%
30
Grok 4.1 Fast (Reasoning)
93.3%
31
Qwen3.7 Plus
93%
32
Hy3 Preview
92.7%
33
GPT-5.2-Codex
92.1%
34
Claude Opus 4.6 (Adaptive)
92.1%
35
Muse Spark
91.5%
36
MiMo-V2-Omni
91.2%
37
Trinity-Large-Preview
90.1%
38
Trinity-Large-Thinking
90.1%
39
Claude Opus 4.5 Thinking
89.5%
40
Qwen3.5-35B-A3B
89.2%
41
MiniMax M3
88.9%
42
Claude Opus 4.7 (Adaptive)
88.6%
43
GPT-5.4
87.1%
44
Gemini 3 Pro
87.1%
45
GPT-5 (medium)
86.5%
46
Claude Opus 4.5
86.3%
47
GPT-5.3 Codex
86%
48
Ling 2.6 Flash
86%
49
Claude Opus 4.6
84.8%
50
GPT-5.2
84.8%
51
GPT-5 (high)
84.8%
52
MiniMax M2.7
84.8%
53
Qwen3.5 397B
83.9%
54
MiMo-V2-Flash
83.9%
55
Nemotron 3 Ultra
83.3%
56
GPT-5.4 mini
83.3%
57
GPT-5.1-Codex-Max
83%
58
GPT-5.1-Codex
83%
59
GPT-5.1
81.9%
60
o3
80.7%
61
Command A+
80.7%
62
Claude Sonnet 4.6
79.5%
63
DeepSeek V3.2
78.9%
64
GLM-4.6
76.9%
65
GPT-5.4 nano
76%
66
Grok Code Fast 1
75.7%
67
Grok 4
74.9%
68
K-Exaone
74.3%
69
Qwen3 Max
74.3%
70
Claude Opus 4.7
74%
71
Claude 4.1 Opus Thinking
71.4%
72
GPT-OSS 120B
65.8%
73
Grok 4 Fast (Reasoning)
65.8%
74
Grok 4.1 Fast
63.7%
75
o1
62.6%
76
Kimi K2
61.1%
77
GPT-OSS 20B
60.2%
78
Gemma 4 31B
59.9%
79
Gemini 2.5 Pro
54.1%
80
GPT-4.1 mini
52.9%
81
Claude 4 Sonnet
52.3%
82
GPT-4.1
47.1%
83
Sarvam 105B
46.8%
84
GLM-4.5-Air
46.5%
85
Nemotron 3 Nano Omni 30B A3B
45.3%
86
Gemma 4 26B A4B
43.6%
87
Gemini 3 Flash
43.3%
88
Mistral Small 4 (Reasoning)
41.2%
89
Mistral Small 4
41.2%
90
DeepSeek V3.1 (Reasoning)
37.4%
91
DeepSeek-R1
36.5%
92
Gemma 4 12B
36.3%
93
DeepSeek V3.1
34.8%
94
Sarvam 30B
34.5%
95
Solar Pro 2
31.9%
96
Gemini 3.1 Flash-Lite
31.3%
97
Mistral Large 2
30.7%
98
o3-mini
28.7%
99
Nemotron 3 Nano 30B
25.4%
100
GPT-4o
25.1%
101
Mistral Large 3
24.6%
102
Mistral Medium 3
24.3%
103
DeepSeek V3
22.8%
104
Granite-4.0-1B
22.8%
105
Claude 3 Haiku
21.1%
106
Gemma 4 E4B
20.8%
107
Gemma 4 E2B
20.8%
108
Exaone 4.0 1.2B
20.5%
109
Granite-4.0-H-1B
19.6%
110
Llama 3.1 405B
19%
111
Llama 4 Maverick
17.8%
112
GPT-4.1 nano
17.3%
113
LFM2.5-8B-A1B
16.1%
114
Llama 4 Scout
15.5%
115
Gemini 2.5 Flash
14.9%
116
Granite-4.0-H-350M
14.6%
117
Nova Pro
14%
118
Granite-4.0-350M
13.2%
119
Nemotron Ultra 253B
11.4%
120
Gemma 3 27B
10.5%
121
LFM2.5-VL-1.6B-Extract
8.5%
122
Exaone 4.0 32B
4.1%
123
Phi-4
0%

FAQ

What does Tau2-Telecom measure?

A telecom-oriented tool benchmark that measures structured tool use in domain workflows.

Which model scores highest on Tau2-Telecom?

Step 3.7 Flash by StepFun currently leads with a score of 98.5% on Tau2-Telecom.

How many models are evaluated on Tau2-Telecom?

123 AI models have been evaluated on Tau2-Telecom on BenchLM.

Compare Top Models on Tau2-Telecom

Step 3.7 Flash vs GLM-5V-TurboGLM-5V-Turbo vs GLM-5-TurboGLM-5-Turbo vs GLM-5GLM-5 vs GLM-5.1
Last updated: June 13, 2026 · BenchLM version τ²-Bench 2026

The AI models change fast. We track them for you.

For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.

Free. No spam. Unsubscribe anytime.