TAU-bench Airline

Agentic

Source ↗ Higher is better

Rank	Model	Provider	Score
1	Anthropic: Claude Opus 4.5	Anthropic	84.0%
2	OpenAI: GPT-5.2	OpenAI	83.0%
3	Gemini 3 Flash	Google	82.5%
4	Z.ai: GLM 5	Z.ai	82.5%
5	Qwen: Qwen3.5 397B A17B	Alibaba	81.5%
6	gemini-3-pro	Google	80.5%
7	Gemini 3.0 Pro	Google	73.0%
8	Anthropic: Claude Sonnet 4.5	Anthropic	72.0%
9	Qwen: Qwen3 Max ThinkingHistorical	Alibaba	69.0%
10	Claude-3.7-Sonnet	Anthropic	64.2%
11	DeepSeek: DeepSeek V3.2Historical	DeepSeek	63.8%
12	OpenAI: GPT-5	OpenAI	62.5%
13	Anthropic: Claude Sonnet 4Historical	Anthropic	60.0%
14	Claude Opus 4	Anthropic	59.6%
15	Qwen: Qwen3 MaxHistorical	Alibaba	59.5%
16	MoonshotAI: Kimi K2 0711	Moonshot AI	56.5%
17	Anthropic: Claude Opus 4.1	Anthropic	56.0%
18	OpenAI: GPT-4.1	OpenAI	56.0%
19	Nemotron-Orchestrator-8B	NVIDIA	56.0%
20	OpenAI: o4 Mini	OpenAI	52.1%
21	OpenAI: o3	OpenAI	52.0%
22	OpenAI: GPT-4.1 Mini	OpenAI	48.7%

Access this data

JSON GET /api/v1

[
  {
    "rank": 1,
    "model": "Anthropic: Claude Opus 4.5",
    "provider": "Anthropic",
    "score": "84.0"
  },
  {
    "rank": 2,
    "model": "OpenAI: GPT-5.2",
    "provider": "OpenAI",
    "score": "83.0"
  },
  {
    "rank": 3,
    "model": "Gemini 3 Flash",
    "provider": "Google",
    "score": "82.5"
  }
]

Curl

$ curl -s https://rightrank.ai/api/v1/rankings/tau-bench-airline.json | jq '.[:3]'

CLI brew install rightrank

$ rightrank top tau-bench-airline --limit 5
$ rightrank compare claude-opus-4.6 gpt-5.4-pro --category tau-bench-airline
$ rightrank value tau-bench-airline  # best performance per dollar

Claude Code Skill Skill

# In Claude Code, use the RightRank skill:
/rightrank best model for tau-bench-airline
/rightrank compare top 5 by value
/rightrank cheapest model scoring above 80 on tau-bench-airline