{"benchmark_id":"legal-agent-benchmark","benchmark_name":"Legal Agent Benchmark","benchmark_description":"The Legal Agent Benchmark (LAB) is Harvey's open-source benchmark for evaluating AI agents on complex, long-horizon legal work. Tasks are scored under an all-pass standard against expert-curated rubrics, where a task passes only if every required rubric criterion (facts, conclusions, citations, structure, and analytical moves) passes.","max_score":1.0,"categories":["legal","reasoning","agents"],"modality":"text","total_models":13,"entries":[{"rank":1,"model_id":"claude-fable-5","model_name":"Claude Fable 5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.133,"normalized_score":0.133,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":10.0,"output_cost_per_million":50.0,"speed_rps":null,"context_window":1000000,"release_date":"2026-06-09","announcement_date":"2026-06-09","multimodal":true,"param_count":null,"is_new":false},{"rank":2,"model_id":"claude-opus-5","model_name":"Claude Opus 5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.117,"normalized_score":0.117,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":5.0,"output_cost_per_million":25.0,"speed_rps":null,"context_window":1000000,"release_date":"2026-07-24","announcement_date":"2026-07-24","multimodal":true,"param_count":null,"is_new":false},{"rank":3,"model_id":"claude-opus-4-7","model_name":"Claude Opus 4.7","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.071,"normalized_score":0.071,"verified":false,"self_reported":false,"provider_id":"anthropic","input_cost_per_million":5.0,"output_cost_per_million":25.0,"speed_rps":42.0,"context_window":1000000,"release_date":"2026-04-16","announcement_date":"2026-04-16","multimodal":true,"param_count":null,"is_new":false},{"rank":4,"model_id":"claude-sonnet-5","model_name":"Claude Sonnet 5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.058,"normalized_score":0.058,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":2.0,"output_cost_per_million":10.0,"speed_rps":42.0,"context_window":1000000,"release_date":"2026-06-30","announcement_date":"2026-06-30","multimodal":true,"param_count":null,"is_new":false},{"rank":5,"model_id":"claude-sonnet-4-6","model_name":"Claude Sonnet 4.6","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.054,"normalized_score":0.054,"verified":false,"self_reported":false,"provider_id":"anthropic","input_cost_per_million":3.0,"output_cost_per_million":15.0,"speed_rps":42.0,"context_window":200000,"release_date":"2026-02-17","announcement_date":"2026-02-17","multimodal":true,"param_count":null,"is_new":false},{"rank":6,"model_id":"claude-opus-4-6","model_name":"Claude Opus 4.6","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.042,"normalized_score":0.042,"verified":false,"self_reported":false,"provider_id":"anthropic","input_cost_per_million":5.0,"output_cost_per_million":25.0,"speed_rps":42.0,"context_window":1000000,"release_date":"2026-02-05","announcement_date":"2026-02-05","multimodal":true,"param_count":null,"is_new":false},{"rank":7,"model_id":"gpt-5.5","model_name":"GPT-5.5","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.021,"normalized_score":0.021,"verified":false,"self_reported":false,"provider_id":"openai","input_cost_per_million":5.0,"output_cost_per_million":30.0,"speed_rps":null,"context_window":1050000,"release_date":"2026-04-23","announcement_date":"2026-04-23","multimodal":true,"param_count":null,"is_new":false},{"rank":8,"model_id":"gemini-3.5-flash","model_name":"Gemini 3.5 Flash","organization_name":"Google","organization_id":"google","benchmark_score":0.008,"normalized_score":0.008,"verified":false,"self_reported":false,"provider_id":"google","input_cost_per_million":1.5,"output_cost_per_million":9.0,"speed_rps":null,"context_window":1048576,"release_date":"2026-05-19","announcement_date":"2026-05-19","multimodal":true,"param_count":null,"is_new":false},{"rank":9,"model_id":"gpt-5.4","model_name":"GPT-5.4","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.004,"normalized_score":0.004,"verified":false,"self_reported":false,"provider_id":"openai","input_cost_per_million":2.5,"output_cost_per_million":15.0,"speed_rps":50.0,"context_window":1000000,"release_date":"2026-03-05","announcement_date":"2026-03-05","multimodal":true,"param_count":null,"is_new":false},{"rank":10,"model_id":"gpt-5.4-mini","model_name":"GPT-5.4 mini","organization_name":"OpenAI","organization_id":"openai","benchmark_score":0.0,"normalized_score":0.0,"verified":false,"self_reported":false,"provider_id":"openai","input_cost_per_million":0.75,"output_cost_per_million":4.5,"speed_rps":null,"context_window":400000,"release_date":"2026-03-17","announcement_date":"2026-03-17","multimodal":true,"param_count":null,"is_new":false},{"rank":11,"model_id":"gemini-3.1-flash-lite-preview","model_name":"Gemini 3.1 Flash-Lite","organization_name":"Google","organization_id":"google","benchmark_score":0.0,"normalized_score":0.0,"verified":false,"self_reported":false,"provider_id":"google","input_cost_per_million":0.25,"output_cost_per_million":1.5,"speed_rps":null,"context_window":1000000,"release_date":"2026-03-03","announcement_date":"2026-03-03","multimodal":true,"param_count":null,"is_new":false},{"rank":12,"model_id":"gemini-3.1-pro-preview","model_name":"Gemini 3.1 Pro","organization_name":"Google","organization_id":"google","benchmark_score":0.0,"normalized_score":0.0,"verified":false,"self_reported":false,"provider_id":"google","input_cost_per_million":2.5,"output_cost_per_million":15.0,"speed_rps":90.0,"context_window":1048576,"release_date":"2026-02-19","announcement_date":"2026-02-19","multimodal":true,"param_count":null,"is_new":false},{"rank":13,"model_id":"gemini-3-flash-preview","model_name":"Gemini 3 Flash","organization_name":"Google","organization_id":"google","benchmark_score":0.0,"normalized_score":0.0,"verified":false,"self_reported":false,"provider_id":"google","input_cost_per_million":0.5,"output_cost_per_million":3.0,"speed_rps":null,"context_window":1000000,"release_date":"2025-12-17","announcement_date":"2025-12-17","multimodal":true,"param_count":null,"is_new":false}]}