Benchmark Radar™
Star —

AI Professions Prompt Benchmark

Explore score history ↗

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.

Released
2026-10-22
Evaluates
professional
Openness
unknown
Importer
Model reports

AI Professions Prompt Benchmark code

AI Professions Prompt Benchmark results and reported scores

Reported benchmark scores · Model reports

Reported scores: 5

ModelOrganizationReported valueReportedEvidence
Claude Fable 5.1Anthropic4.14322026-10-22evidence
Claude Opus 5.5Anthropic4.29692026-10-22evidence
Claude Sonnet 5Anthropic3.97982026-10-22evidence
Gemini 4 Argon MediumGoogle4.24352026-10-22evidence
GPT-6 AstraOpenAI4.87662026-10-22evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.

Which sources cite AI Professions Prompt Benchmark?

Related professional benchmarks