Enter opens the highlighted result. Esc closes.

JobBench leaderboard

As of November 14, 2026, among the results listed here, Meta Muse Wonder has the highest reported JobBench score, 70.5 (max), published by Meta. 19 models have a JobBench result in our records.

Scores as the model makers published them, not re-run by us; each row links to its announcement.

Best reported score

70.5

Meta Muse Wonder

Reported by Meta, October 12, 2026 · max

Tied for first

70.5

T Thinking Machines Inkling 2 Small

Reported by Thinking Machines, November 9, 2026

19 models ranked · 7 makers

Reported JobBench scores

JobBench scores reported by model makers
Rank Model Score Setting Reported by Date
1 Muse Wonder 70.5 max Meta October 12, 2026 (model release)
1 T Inkling 2 Small 70.5 Not published Thinking Machines November 9, 2026 (model release)
3 Muse Spark 1.3 64.9 max Meta August 5, 2026 (model release)
4 MiMo-V2.6-Pro 62 Not published Xiaomi September 21, 2026 (model release)
5 Muse Spark Open 61.8 xhigh Meta October 15, 2026 (model release)
6 Muse Spark 1.2 61.6 xhigh Meta August 5, 2026 (model release)
7 MiMo-V2.6-Flash 61.2 Not published Xiaomi September 21, 2026 (model release)
8 Step 5 Preview 59 High StepFun October 8, 2026 (model release)
9 Qwen3.8 Flash 55.7 Not published Qwen August 26, 2026 (model release)
10 Muse Spark 1.1 54.7 Not published Meta July 16, 2026 (model release)
11 Kimi K3 54.3 max Moonshot AI July 16, 2026 (model release)
12 Qwen3.8 Max (0902) 53.4 Not published Qwen August 3, 2026
13 N Nex-N2.5-Pro 41.4 Not published Nex AGI September 8, 2026 (model release)
14 Qwen3.8 27B 33.4 Not published Qwen August 14, 2026 (model release)
15 Qwen3.7 Max 31.3 Not published Qwen August 3, 2026
16 N Nex-N2.5-Mini 28.5 Not published Nex AGI September 8, 2026 (model release)
17 Qwen3.7 Plus 27.6 Not published Qwen August 14, 2026 (model release)
18 MiMo-V2.5-Pro 25 Not published Xiaomi September 21, 2026 (model release)
19 Qwen3.6 27B 21.8 Not published Qwen August 14, 2026 (model release)

How much to trust these numbers

Source
Reported by the model makers themselves. Not re-run by LLMBenchmark or anyone independent.
Settings
Settings differ between makers. Check the Setting column before comparing two scores.
Spread
70.5 to 21.8 across 19 models.
Freshness
Newest result published November 9, 2026; oldest July 16, 2026.

What JobBench measures

JobBench evaluates how well AI agents perform workflows experts prioritize for delegation, with scores reflecting performance on those workflows. Source: JobBench: Aligning Agent Work With Human Will