Professional tool use

JobBench

Tests the AI on professional workplace tasks that require using real work tools — the kind of multi-step jobs an office worker handles. Higher is better.

Rankings

Higher is better

JobBench — frequently asked questions

What is JobBench?
Tests the AI on professional workplace tasks that require using real work tools — the kind of multi-step jobs an office worker handles. Higher is better.
Which AI model scores highest on JobBench?
Muse Spark 1.1 by Meta holds the best JobBench result among tracked models, at 54.7% (released Jul 9 2026). Higher scores are better on this benchmark.
What are the top 5 models on JobBench?
1. Muse Spark 1.1 (Meta) — 54.7%; 2. Qwen3.8-Max (Qwen) — 53.4%; 3. Kimi K3 (Moonshot AI) — 52.9%; 4. Qwen3.8-27B (Qwen) — 33.4%; 5. Muse Spark (Meta) — 17%.
How many models have a published JobBench score?
5 tracked models have a published JobBench score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.
What is the best open model on JobBench?
Qwen3.8-27B by Qwen is the highest-ranked model with downloadable weights on JobBench, scoring 33.4% at rank 4 overall.