{"record": {"aliases": ["AI Professions Prompt Benchmark", "aiprofessions/prompt-benchmark", "AI Professions prompt benchmark: profession prompts, hash-locked rubrics and graded LLM runs"], "artifacts": [{"kind": "repo", "url": "https://github.com/aiprofessions/prompt-benchmark"}], "audits": [], "categories": ["professional"], "description": {"en": "100 prompts with fictional inputs, five pass/fail criteria each, scored as the share of criteria met over three plain-chat runs without tools. Claude and Gemini outputs are graded by gpt-6-astra and gpt-6-astra outputs by claude-opus-5-5; the README warns the two judges differ in strictness, so only models graded by the same judge compare directly. Results grow in rounds, so the model set depends on the release read."}, "documents": [{"document_type": "benchmark_leaderboard", "id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "model_name": null, "organization": "AI Professions", "published": "2026-10-22", "retrieved_at": "2026-11-19", "source": "model_reports", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)"}], "evidence_summary": {"document_count": 1, "model_count": 5, "model_count_basis": "source_model_id"}, "key": "model-reports:ai_professions_prompt_benchmark", "modality": null, "name": "AI Professions Prompt Benchmark", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"registry": "data/model_cards.yml", "score_archive": "data/benchmark_scores.yml", "source_url": "https://github.com/aiprofessions/prompt-benchmark"}, "publisher": null, "released": "2026-10-22", "released_reference": {"basis": "benchmark_release", "source_key": "model-reports:ai_professions_prompt_benchmark", "source_url": "https://github.com/aiprofessions/prompt-benchmark"}, "schema_version": 1, "sizes": [], "slug": "ai_professions_prompt_benchmark", "source": "model_reports", "source_benchmark_id": "ai_professions_prompt_benchmark"}, "schema_version": 1, "scores_by_source": {"model_reports": {"rows": [{"benchmark_id": "ai_professions_prompt_benchmark", "comparable_group": "[\"model-reports:ai_professions_prompt_benchmark\",\"ai_professions_prompt_benchmark_6d4c6d8b\",\"100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\"]", "date_precision": "document_publication", "document_id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "instrument": "ai_professions_prompt_benchmark_6d4c6d8b", "key": "model-reports:ai_professions_prompt_benchmark", "measured_by": "AI Professions", "measurement_kind": "benchmark_publisher_run", "model": "GPT-6 Astra", "model_id": "[\"OpenAI\",\"GPT-6 Astra\"]", "model_name": "GPT-6 Astra", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "organization": "OpenAI", "protocol": "100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5", "rank_in_source_response": null, "raw_value": "4.8766", "read_from": "html_text", "reported_at": "2026-10-22", "reported_by": "third_party", "reported_date": "2026-10-22", "series_id": "model_reports:ai_professions_prompt_benchmark:default", "source": "model_reports", "source_document_type": "benchmark_leaderboard", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "value": 4.8766, "value_kind": "number"}, {"benchmark_id": "ai_professions_prompt_benchmark", "comparable_group": "[\"model-reports:ai_professions_prompt_benchmark\",\"ai_professions_prompt_benchmark_6d4c6d8b\",\"100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\"]", "date_precision": "document_publication", "document_id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "instrument": "ai_professions_prompt_benchmark_6d4c6d8b", "key": "model-reports:ai_professions_prompt_benchmark", "measured_by": "AI Professions", "measurement_kind": "benchmark_publisher_run", "model": "Claude Fable 5.1", "model_id": "[\"Anthropic\",\"Claude Fable 5.1\"]", "model_name": "Claude Fable 5.1", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Fable 5.1", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Fable 5.1", "organization": "Anthropic", "protocol": "100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra", "rank_in_source_response": null, "raw_value": "4.1432", "read_from": "html_text", "reported_at": "2026-10-22", "reported_by": "third_party", "reported_date": "2026-10-22", "series_id": "model_reports:ai_professions_prompt_benchmark:default", "source": "model_reports", "source_document_type": "benchmark_leaderboard", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "value": 4.1432, "value_kind": "number"}, {"benchmark_id": "ai_professions_prompt_benchmark", "comparable_group": "[\"model-reports:ai_professions_prompt_benchmark\",\"ai_professions_prompt_benchmark_6d4c6d8b\",\"100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\"]", "date_precision": "document_publication", "document_id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "instrument": "ai_professions_prompt_benchmark_6d4c6d8b", "key": "model-reports:ai_professions_prompt_benchmark", "measured_by": "AI Professions", "measurement_kind": "benchmark_publisher_run", "model": "Claude Opus 5.5", "model_id": "[\"Anthropic\",\"Claude Opus 5.5\"]", "model_name": "Claude Opus 5.5", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Opus 5.5", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Opus 5.5", "organization": "Anthropic", "protocol": "100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra", "rank_in_source_response": null, "raw_value": "4.2969", "read_from": "html_text", "reported_at": "2026-10-22", "reported_by": "third_party", "reported_date": "2026-10-22", "series_id": "model_reports:ai_professions_prompt_benchmark:default", "source": "model_reports", "source_document_type": "benchmark_leaderboard", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "value": 4.2969, "value_kind": "number"}, {"benchmark_id": "ai_professions_prompt_benchmark", "comparable_group": "[\"model-reports:ai_professions_prompt_benchmark\",\"ai_professions_prompt_benchmark_6d4c6d8b\",\"100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\"]", "date_precision": "document_publication", "document_id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "instrument": "ai_professions_prompt_benchmark_6d4c6d8b", "key": "model-reports:ai_professions_prompt_benchmark", "measured_by": "AI Professions", "measurement_kind": "benchmark_publisher_run", "model": "Claude Sonnet 5", "model_id": "[\"Anthropic\",\"Claude Sonnet 5\"]", "model_name": "Claude Sonnet 5", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Sonnet 5", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Claude Sonnet 5", "organization": "Anthropic", "protocol": "100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra", "rank_in_source_response": null, "raw_value": "3.9798", "read_from": "html_text", "reported_at": "2026-10-22", "reported_by": "third_party", "reported_date": "2026-10-22", "series_id": "model_reports:ai_professions_prompt_benchmark:default", "source": "model_reports", "source_document_type": "benchmark_leaderboard", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "value": 3.9798, "value_kind": "number"}, {"benchmark_id": "ai_professions_prompt_benchmark", "comparable_group": "[\"model-reports:ai_professions_prompt_benchmark\",\"ai_professions_prompt_benchmark_6d4c6d8b\",\"100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\"]", "date_precision": "document_publication", "document_id": "model_reports:ai_professions_prompt_benchmark_results_2026_10_22", "instrument": "ai_professions_prompt_benchmark_6d4c6d8b", "key": "model-reports:ai_professions_prompt_benchmark", "measured_by": "AI Professions", "measurement_kind": "benchmark_publisher_run", "model": "Gemini 4 Argon Medium", "model_id": "[\"Google\",\"Gemini 4 Argon Medium\"]", "model_name": "Gemini 4 Argon Medium", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Gemini 4 Argon Medium", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra\u0000Gemini 4 Argon Medium", "organization": "Google", "protocol": "100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by gpt-6-astra", "rank_in_source_response": null, "raw_value": "4.2435", "read_from": "html_text", "reported_at": "2026-10-22", "reported_by": "third_party", "reported_date": "2026-10-22", "series_id": "model_reports:ai_professions_prompt_benchmark:default", "source": "model_reports", "source_document_type": "benchmark_leaderboard", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_title": "AI Professions prompt benchmark results (pinned commit 6d4c6d8b, runs 2026-10-19 to 2026-10-21)", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv", "value": 4.2435, "value_kind": "number"}], "series": {"declared_max": null, "direction": "higher_is_better", "direction_basis": "reported_metric", "display_scale": null, "first_score_record": {"date_precision": "document_publication", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "reported_at": "2026-10-22", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv"}, "first_score_report": {"date_precision": "document_publication", "obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "reported_at": "2026-10-22", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv"}, "key": "model-reports:ai_professions_prompt_benchmark", "max_score_contradicted": false, "metric": "criteria_passed_avg", "observation_count": 5, "observed_max": 4.8766, "score_summary": {"display_max": 4.8766, "display_multiplier": 1, "model_count": 5, "model_count_basis": "source_model_id", "numeric_count": 5, "raw_max": 4.8766, "raw_min": 3.9798, "source_reference": {"obs_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "observation_id": "curated\u0000ai_professions_prompt_benchmark\u0000ai_professions_prompt_benchmark_results_2026_10_22\u0000ai_professions_prompt_benchmark_6d4c6d8b\u0000100 prompts, 3 plain-chat runs each, 5 pass/fail criteria per run, no tools; judged by claude-opus-5-5\u0000GPT-6 Astra", "reported_at": "2026-10-22", "reported_date": "2026-10-22", "source_id": "ai_professions_prompt_benchmark_results_2026_10_22", "source_url": "https://github.com/aiprofessions/prompt-benchmark/blob/6d4c6d8b28756d23024b6f4143f1a8938160278a/data/results.csv"}, "unit": "criteria"}, "series_id": "model_reports:ai_professions_prompt_benchmark:default", "unit": "criteria"}}}, "siblings": []}
