{"benchmark_id":"program-bench","benchmark_name":"Program Bench","benchmark_description":"Program Bench evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and documentation. It spans 200 tasks from small CLI tools to large systems such as FFmpeg and SQLite, with submissions judged against more than 248,000 fuzz-generated behavioral tests.","max_score":1.0,"categories":["agents","code"],"modality":"text","total_models":15,"entries":[{"rank":1,"model_id":"claude-opus-5-5","model_name":"Claude Opus 5.5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.912,"normalized_score":0.912,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":4.0,"output_cost_per_million":20.0,"speed_rps":null,"context_window":1000000,"release_date":"2026-09-22","announcement_date":"2026-09-22","multimodal":true,"param_count":null,"is_new":false},{"rank":2,"model_id":"claude-haiku-5-5","model_name":"Claude Haiku 5.5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.82,"normalized_score":0.82,"verified":false,"self_reported":true,"provider_id":"anthropic","input_cost_per_million":0.1,"output_cost_per_million":0.5,"speed_rps":null,"context_window":1000000,"release_date":"2026-10-07","announcement_date":"2026-10-07","multimodal":true,"param_count":null,"is_new":true},{"rank":3,"model_id":"claude-sonnet-5-5","model_name":"Claude Sonnet 5.5","organization_name":"Anthropic","organization_id":"anthropic","benchmark_score":0.797,"normalized_score":0.797,"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":"2026-09-28","announcement_date":"2026-09-28","multimodal":true,"param_count":null,"is_new":false},{"rank":4,"model_id":"kimi-k3","model_name":"Kimi K3","organization_name":"Moonshot AI","organization_id":"moonshotai","benchmark_score":0.778,"normalized_score":0.778,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":2.85,"output_cost_per_million":14.25,"speed_rps":null,"context_window":1048576,"release_date":"2026-07-16","announcement_date":"2026-07-16","multimodal":true,"param_count":2800000000000,"is_new":false},{"rank":5,"model_id":"glm-5.2","model_name":"GLM-5.2","organization_name":"Zhipu AI","organization_id":"zai-org","benchmark_score":0.637,"normalized_score":0.637,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":0.75,"output_cost_per_million":2.4,"speed_rps":null,"context_window":1048576,"release_date":"2026-06-16","announcement_date":"2026-06-16","multimodal":false,"param_count":753000000000,"is_new":false},{"rank":6,"model_id":"muse-wonder-full","model_name":"Muse Wonder Full","organization_name":"Meta","organization_id":"meta","benchmark_score":0.59,"normalized_score":0.59,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-10-12","announcement_date":"2026-10-12","multimodal":true,"param_count":null,"is_new":true},{"rank":7,"model_id":"kimi-k2.7-code","model_name":"Kimi K2.7 Code","organization_name":"Moonshot AI","organization_id":"moonshotai","benchmark_score":0.536,"normalized_score":0.536,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":0.68,"output_cost_per_million":3.4,"speed_rps":null,"context_window":262144,"release_date":"2026-06-12","announcement_date":"2026-06-12","multimodal":true,"param_count":1000000000000,"is_new":false},{"rank":8,"model_id":"seed-2.1-pro","model_name":"Seed 2.1 Pro","organization_name":"ByteDance","organization_id":"bytedance","benchmark_score":0.503,"normalized_score":0.503,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-06-24","announcement_date":"2026-06-24","multimodal":true,"param_count":null,"is_new":false},{"rank":9,"model_id":"muse-wonder","model_name":"Muse Wonder","organization_name":"Meta","organization_id":"meta","benchmark_score":0.5,"normalized_score":0.5,"verified":false,"self_reported":true,"provider_id":"meta","input_cost_per_million":4.0,"output_cost_per_million":20.0,"speed_rps":null,"context_window":1048576,"release_date":"2026-10-12","announcement_date":"2026-10-12","multimodal":true,"param_count":null,"is_new":true},{"rank":10,"model_id":"seed-2.1-turbo","model_name":"Seed 2.1 Turbo","organization_name":"ByteDance","organization_id":"bytedance","benchmark_score":0.494,"normalized_score":0.494,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-06-24","announcement_date":"2026-06-24","multimodal":true,"param_count":null,"is_new":false},{"rank":11,"model_id":"mimo-v2.6-pro","model_name":"MiMo-V2.6-Pro","organization_name":"Xiaomi","organization_id":"xiaomi","benchmark_score":0.265,"normalized_score":0.265,"verified":false,"self_reported":true,"provider_id":"xiaomi","input_cost_per_million":0.435,"output_cost_per_million":0.87,"speed_rps":null,"context_window":1048576,"release_date":"2026-09-22","announcement_date":"2026-09-22","multimodal":true,"param_count":1020000000000,"is_new":false},{"rank":12,"model_id":"mimo-v2.6-flash","model_name":"MiMo-V2.6-Flash","organization_name":"Xiaomi","organization_id":"xiaomi","benchmark_score":0.26,"normalized_score":0.26,"verified":false,"self_reported":true,"provider_id":"xiaomi","input_cost_per_million":0.14,"output_cost_per_million":0.28,"speed_rps":null,"context_window":1048576,"release_date":"2026-09-22","announcement_date":"2026-09-22","multimodal":true,"param_count":309000000000,"is_new":false},{"rank":13,"model_id":"deepseek-v4.1-flash","model_name":"DeepSeek-V4.1-Flash","organization_name":"DeepSeek","organization_id":"deepseek","benchmark_score":0.203,"normalized_score":0.203,"verified":false,"self_reported":true,"provider_id":"fireworks","input_cost_per_million":0.22,"output_cost_per_million":0.66,"speed_rps":null,"context_window":1040000,"release_date":"2026-09-10","announcement_date":"2026-09-10","multimodal":true,"param_count":763205315794,"is_new":false},{"rank":14,"model_id":"glm-5.3","model_name":"GLM-5.3","organization_name":"Zhipu AI","organization_id":"zai-org","benchmark_score":0.19,"normalized_score":0.19,"verified":false,"self_reported":true,"provider_id":"deepinfra","input_cost_per_million":1.2,"output_cost_per_million":4.0,"speed_rps":null,"context_window":1048576,"release_date":"2026-08-14","announcement_date":"2026-08-14","multimodal":false,"param_count":753000000000,"is_new":false},{"rank":15,"model_id":"hy4-preview","model_name":"Hy4 preview","organization_name":"Tencent","organization_id":"tencent","benchmark_score":0.175,"normalized_score":0.175,"verified":false,"self_reported":true,"provider_id":null,"input_cost_per_million":null,"output_cost_per_million":null,"speed_rps":null,"context_window":null,"release_date":"2026-08-28","announcement_date":"2026-08-28","multimodal":false,"param_count":770000000000,"is_new":false}]}