Agentic benchmark
JobBench leaderboard
Every model the catalog carries a published JobBench value for, ranked by that value.
CategoryAgentic
MeasureAgentic workplace deliverables
Tasks130 tasks across 35 occupations
DifficultyProfessional multi-source workflows
Published byJobBench: Aligning Agent Work With Human Will
An occupational agent benchmark for professional workflows that workers say they most want delegated to AI.
JobBench ranking
27 models with a published JobBench value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Agentic workplace deliverables |
|---|---|---|---|
| 1 |  Muse Spark 1.3 | Meta | 64.9 |
| 2 |  Hy4 preview | Tencent | 61.7 |
| 3 |  Qwen3.8-Flash-Next | Alibaba | 55.7 |
| 4 | Muse Spark 1.1 | Meta | 54.7 |
| 5 | Qwen3.8 Max | Alibaba | 53.4 |
| 6 |  Kimi K3 | Moonshot AI | 52.9 |
| 7 |  Atria Dawn Preview | Shanghai Artificial Intelligence Laboratory | 50.3 |
| 8 |  Claude Opus 4.7 (Adaptive) | Anthropic | 45.9 |
| 9 |  GPT-5.5 | OpenAI | 42.7 |
| 10 | GPT-5.4 | OpenAI | 38.9 |
| 11 | Claude Sonnet 4.6 | Anthropic | 36.9 |
| 12 | Claude Opus 4.6 | Anthropic | 36.7 |
| 13 | GPT-5.2 | OpenAI | 34.3 |
| 14 | GPT-5.3 Codex | OpenAI | 33.7 |
| 15 | Qwen3.8-27B | Alibaba | 33.4 |
| 16 | Claude Opus 4.5 | Anthropic | 32.3 |
| 17 | Claude Sonnet 4.5 | Anthropic | 27.7 |
| 18 | GPT-5.1-Codex | OpenAI | 26.2 |
| 19 | GPT-5.2-Codex | OpenAI | 26.0 |
| 20 | Claude 4.1 Opus | Anthropic | 21.9 |
| 21 | Qwen3.5 Plus | Alibaba | 18.5 |
| 22 | Claude 4 Sonnet | Anthropic | 18.4 |
| 23 | Claude Haiku 4.5 | Anthropic | 16.0 |
| 24 |  Gemini 3 Flash | 11.4 | |
| 24 | Gemini 3 Pro | 11.4 | |
| 26 | Kimi K2.5 | Moonshot AI | 8.7 |
| 27 | GPT-5 (high) | OpenAI | 8.5 |
Evidence key: Observed
Rows are ordered by the value JobBench: Aligning Agent Work With Human Will published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.