Atlas

Benchmarks

OpenWorldEngineer — ML

Mercor

OpenWorldEngineer Mean@8 on its 7 ML tasks (kernels, training and inference: attention on B300 and H100, wave-propagation stencils, post-training, context extension, representation learning and speculative decoding).

Models
15
Top score
54.6
Leader
Claude Fable 5.2
Updated
—
Resolution
—
Saturation
—
Signal
95thpercentile
192 benchmarks from 0.8 to 3.5218 benchmarks from 3.5 to 6.1128 benchmarks from 6.1 to 8.895 benchmarks from 8.8 to 11.567 benchmarks from 11.5 to 14.128 benchmarks from 14.1 to 16.814 benchmarks from 16.8 to 19.49 benchmarks from 19.4 to 22.111 benchmarks from 22.1 to 24.72 benchmarks from 24.7 to 27.46 benchmarks from 27.4 to 30.13 benchmarks from 30.1 to 32.73 benchmarks from 32.7 to 35.45 benchmarks from 35.4 to 38.01 benchmark from 38.0 to 40.71 benchmark from 40.7 to 43.41 benchmark from 43.4 to 46.01 benchmark from 62.0 to 64.6

42nd strongest of 785

Signal-to-noise +19.67 for the OpenWorldEngineer (Mean@8) family

Difficulty
Not estimated

Another benchmark represents this family in the fit.

Organizations
15 dated models · 7 frontier modelsClosed weightsOpen weights
Data: Mercoratlas.village.dev
#ModelSettingScoreSource
1Claude Fable 5.2max · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
54.6
OpenWorldEngineer official leaderboard
2Claude Sonnet 5.5max · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
47.8
OpenWorldEngineer official leaderboard
3Claude Opus 5.5max · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
42.6
OpenWorldEngineer official leaderboard
4GPT-6.1 Solmax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
39.4
OpenWorldEngineer official leaderboard
5GPT-6 Astramax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
38.9
OpenWorldEngineer official leaderboard
6Gemini 4 Carbonhigh · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
36.4
OpenWorldEngineer official leaderboard
7Gemini 4 Argonhigh · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
32.2
OpenWorldEngineer official leaderboard
8Claude Fable 5.1max · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
31.9
OpenWorldEngineer official leaderboard
9DeepSeek V4.1 Flashmax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
16.8
OpenWorldEngineer official leaderboard
10Kimi K3.1 Promax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
16.2
OpenWorldEngineer official leaderboard
11GLM-5.3max · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
15.7
OpenWorldEngineer official leaderboard
12Muse Wondermax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
11.9
OpenWorldEngineer official leaderboard
13Grok 4.7x-high · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
6.6
OpenWorldEngineer official leaderboard
14Beam 1.0 501B A23Bmax · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
2.8
OpenWorldEngineer official leaderboard
15Inklingnot reported · OpenWorld harness (OpenWorldEngineer runner) on Harbor 0.23 with Modal sandboxes; web search over an October 1, 2026 web snapshot as the only network access; task time budgets of 4 to 18 hours
0.7
OpenWorldEngineer official leaderboard
Loading Atlas data…