{"record": {"aliases": [], "artifacts": [], "categories": ["safety", "agents", "code"], "description": {"en": "ExploitBench is a cybersecurity benchmark that evaluates a model's ability to discover and exploit software vulnerabilities, reported as the fraction of challenges where the model captures the target (Cap%)."}, "documents": [{"document_type": "registry_page", "id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "title": "ExploitBench"}], "evidence_summary": {"document_count": 1, "model_count": 5, "model_count_basis": "source_model_id"}, "key": "llm-stats:exploitbench", "modality": "text", "name": "ExploitBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"crawl_bundle": "llm_stats_2026-08-17", "crawled_at": "2026-08-17T18:49:35.489925+00:00", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500"}, "publisher": null, "released": null, "released_reference": null, "schema_version": 1, "sizes": [], "slug": "llm-stats-exploitbench", "source": "llm_stats", "source_benchmark_id": "exploitbench"}, "schema_version": 1, "scores_by_source": {"llm_stats": {"rows": [{"comparable_group": null, "crawled_at": "2026-08-17T18:49:35.489925+00:00", "date_precision": "model_announcement", "document_id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "key": "llm-stats:exploitbench", "measured_by": null, "model_id": "claude-fable-5", "model_name": "Claude Fable 5", "obs_id": "llm_stats:exploitbench:claude-fable-5", "organization": "Anthropic", "rank_in_source_response": 1, "raw_value": "0.78", "reported_by": "self_reported", "reported_date": "2026-06-09", "series_id": "llm_stats:exploitbench:default", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "value": 0.78, "value_kind": "number"}, {"comparable_group": null, "crawled_at": "2026-08-17T18:49:35.489925+00:00", "date_precision": "model_announcement", "document_id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "key": "llm-stats:exploitbench", "measured_by": null, "model_id": "gpt-5.6-sol", "model_name": "GPT-5.6 Sol", "obs_id": "llm_stats:exploitbench:gpt-5.6-sol", "organization": "OpenAI", "rank_in_source_response": 2, "raw_value": "0.735", "reported_by": "self_reported", "reported_date": "2026-07-09", "series_id": "llm_stats:exploitbench:default", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "value": 0.735, "value_kind": "number"}, {"comparable_group": null, "crawled_at": "2026-08-17T18:49:35.489925+00:00", "date_precision": "model_announcement", "document_id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "key": "llm-stats:exploitbench", "measured_by": null, "model_id": "glm-5.3", "model_name": "GLM-5.3", "obs_id": "llm_stats:exploitbench:glm-5.3", "organization": "Z.ai", "rank_in_source_response": 3, "raw_value": "0.544", "reported_by": "self_reported", "reported_date": "2026-08-14", "series_id": "llm_stats:exploitbench:default", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "value": 0.544, "value_kind": "number"}, {"comparable_group": null, "crawled_at": "2026-08-17T18:49:35.489925+00:00", "date_precision": "model_announcement", "document_id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "key": "llm-stats:exploitbench", "measured_by": null, "model_id": "gpt-5.6-terra", "model_name": "GPT-5.6 Terra", "obs_id": "llm_stats:exploitbench:gpt-5.6-terra", "organization": "OpenAI", "rank_in_source_response": 4, "raw_value": "0.529", "reported_by": "self_reported", "reported_date": "2026-07-09", "series_id": "llm_stats:exploitbench:default", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "value": 0.529, "value_kind": "number"}, {"comparable_group": null, "crawled_at": "2026-08-17T18:49:35.489925+00:00", "date_precision": "model_announcement", "document_id": "llm_stats:https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "key": "llm-stats:exploitbench", "measured_by": null, "model_id": "gpt-5.6-luna", "model_name": "GPT-5.6 Luna", "obs_id": "llm_stats:exploitbench:gpt-5.6-luna", "organization": "OpenAI", "rank_in_source_response": 5, "raw_value": "0.332", "reported_by": "self_reported", "reported_date": "2026-07-09", "series_id": "llm_stats:exploitbench:default", "source": "llm_stats", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500", "value": 0.332, "value_kind": "number"}], "series": {"bounds": {"basis": "aggregator_declared", "max": 1.0, "min": null}, "declared_max": 1.0, "direction": "higher_is_better", "direction_basis": "source_rank_descending", "display_scale": null, "first_score_record": {"date_precision": "model_announcement", "obs_id": "llm_stats:exploitbench:claude-fable-5", "reported_at": "2026-06-09", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500"}, "first_score_report": null, "key": "llm-stats:exploitbench", "max_score_contradicted": false, "metric": null, "observation_count": 5, "observed_max": 0.78, "score_summary": {"display_max": 78.0, "display_multiplier": 100, "model_count": 5, "model_count_basis": "source_model_id", "numeric_count": 5, "raw_max": 0.78, "raw_min": 0.332, "source_reference": {"crawled_at": "2026-08-17T18:49:35.489925+00:00", "obs_id": "llm_stats:exploitbench:claude-fable-5", "reported_date": "2026-06-09", "source_url": "https://api.zeroeval.com/leaderboard/benchmarks/exploitbench?top_n=500"}, "unit": null}, "series_id": "llm_stats:exploitbench:default"}}}, "siblings": []}
