{"benchmark": {"slug": "hal-tau-bench-airline", "name": "TAU-bench Airline (HAL)", "description": "Customer-service airline booking benchmark on HAL.", "category": "eval", "split": null, "language": null, "task_count": 50, "homepage_url": "https://hal.cs.princeton.edu/taubench_airline", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/taubench_airline", "is_public_leaderboard": true, "harness_count": 2, "model_count": 13, "score_count": 24, "best_success_rate": 56.0}, "protocol": {"eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "temperature": null, "max_turns": null, "context_budget_tokens": null, "n": 24}, "rankings": {"benchmark": {"slug": "hal-tau-bench-airline", "name": "TAU-bench Airline (HAL)", "description": "Customer-service airline booking benchmark on HAL.", "category": "eval", "split": null, "language": null, "task_count": 50, "homepage_url": "https://hal.cs.princeton.edu/taubench_airline", "paper_url": "https://arxiv.org/abs/2510.11977", "hf_dataset_id": null, "official_leaderboard_url": "https://hal.cs.princeton.edu/taubench_airline", "is_public_leaderboard": true, "harness_count": null, "model_count": null, "score_count": null, "best_success_rate": null}, "sort": "success_rate", "rows": [{"rank": 1, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "vendor": "Anthropic"}, "success_rate": 56.0, "resolved_count": null, "total_count": null, "cost_usd": 0.8422, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 2, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o3", "name": "o3", "vendor": "openai"}, "success_rate": 54.0, "resolved_count": null, "total_count": null, "cost_usd": 0.2912, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 3, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus", "name": "Claude 4.1 Opus", "vendor": "Anthropic"}, "success_rate": 54.0, "resolved_count": null, "total_count": null, "cost_usd": 3.6098, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 4, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet-high", "name": "Claude 3.7 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 52.0, "resolved_count": null, "total_count": null, "cost_usd": 0.6388, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 5, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "success_rate": 52.0, "resolved_count": null, "total_count": null, "cost_usd": 2.9996, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 6, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus", "name": "Claude 4.1 Opus", "vendor": "Anthropic"}, "success_rate": 50.0, "resolved_count": null, "total_count": null, "cost_usd": 1.3956, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 7, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-5-medium", "name": "GPT-5 (medium)", "vendor": "OpenAI"}, "success_rate": 48.0, "resolved_count": null, "total_count": null, "cost_usd": 0.4766, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 8, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-v3-march-2025", "name": "DeepSeek V3 (March 2025)", "vendor": "DeepSeek"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 0.1086, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 9, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet", "name": "Claude 3.7 Sonnet", "vendor": "Anthropic"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 0.309, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 10, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-3-7-sonnet-high", "name": "Claude 3.7 Sonnet (high)", "vendor": "Anthropic"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 0.6916, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 11, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-opus", "name": "Claude 4 Opus", "vendor": "Anthropic"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 3.003, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 12, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-opus-high", "name": "Claude 4 Opus (high)", "vendor": "Anthropic"}, "success_rate": 44.0, "resolved_count": null, "total_count": null, "cost_usd": 3.0058, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 13, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o4-mini", "name": "o4 mini", "vendor": "openai"}, "success_rate": 36.0, "resolved_count": null, "total_count": null, "cost_usd": 0.1428, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 14, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-4-1", "name": "GPT-4.1", "vendor": "openai"}, "success_rate": 36.0, "resolved_count": null, "total_count": null, "cost_usd": 0.1636, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 15, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-r1-january-2025", "name": "DeepSeek R1 (January 2025)", "vendor": "DeepSeek"}, "success_rate": 36.0, "resolved_count": null, "total_count": null, "cost_usd": 0.266, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 16, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "claude-4-1-opus-high", "name": "Claude 4.1 Opus (high)", "vendor": "Anthropic"}, "success_rate": 32.0, "resolved_count": null, "total_count": null, "cost_usd": 2.8056, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 17, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-5-medium", "name": "GPT-5 (medium)", "vendor": "OpenAI"}, "success_rate": 30.0, "resolved_count": null, "total_count": null, "cost_usd": 1.0556, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 18, "harness": {"slug": "tau-bench-tool-calling", "name": "TAU-bench Tool Calling", "vendor": null, "scaffold_code": null, "repo_url": null, "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "vendor": "Google"}, "success_rate": 28.0, "resolved_count": null, "total_count": null, "cost_usd": 0.0062, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 19, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gemini-2-0-flash", "name": "Gemini 2.0 Flash", "vendor": "Google"}, "success_rate": 22.0, "resolved_count": null, "total_count": null, "cost_usd": 0.04, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 20, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o3", "name": "o3", "vendor": "openai"}, "success_rate": 20.0, "resolved_count": null, "total_count": null, "cost_usd": 0.9006, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 21, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "o4-mini", "name": "o4 mini", "vendor": "openai"}, "success_rate": 18.0, "resolved_count": null, "total_count": null, "cost_usd": 0.4114, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 22, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-v3-march-2025", "name": "DeepSeek V3 (March 2025)", "vendor": "DeepSeek"}, "success_rate": 18.0, "resolved_count": null, "total_count": null, "cost_usd": 0.2146, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 23, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "gpt-4-1", "name": "GPT-4.1", "vendor": "openai"}, "success_rate": 16.0, "resolved_count": null, "total_count": null, "cost_usd": 0.357, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}, {"rank": 24, "harness": {"slug": "hal-generalist-agent", "name": "HAL Generalist Agent", "vendor": "Princeton PLI", "scaffold_code": null, "repo_url": "https://github.com/princeton-pli/hal-harness", "homepage_url": "https://hal.cs.princeton.edu/", "openrouter_icon_url": null, "leaderboard_icon_url": null}, "model": {"slug": "deepseek-r1-january-2025", "name": "DeepSeek R1 (January 2025)", "vendor": "DeepSeek"}, "success_rate": 10.0, "resolved_count": null, "total_count": null, "cost_usd": 0.6036, "input_tokens": null, "output_tokens": null, "latency_ms": null, "temperature": null, "max_turns": null, "context_budget_tokens": null, "is_harnessrl_measured": false, "eval_protocol": "HAL TAU-bench Airline (HAL) accuracy", "source": "hal_site", "source_url": "https://hal.cs.princeton.edu/taubench_airline", "observed_at": null}]}}