Atlas

Benchmarks

MaxxBench

Unlap

MaxxBench (unLAP) changes the instructions or starting code of 89 DeepSWE v1.1 tasks while keeping the engineering problem, with reference solutions and tests revised to match, to test whether a coding agent's DeepSWE score holds when the requirements shift. Each model runs twice on every task in mini-swe-agent under Pier, with no network but its model API, up to three hours per attempt. The score is the share of the 178 attempts in which every required feature and regression test passes.

Models
12
Top score
76.4
Leader
Claude Opus 5.5
Updated
Nov 7, 2026
Resolution
~2 mo
Sharp
Saturation
~Sep 2027
Projected
Signal
84thpercentile
205 benchmarks from 0.8 to 3.6219 benchmarks from 3.6 to 6.4134 benchmarks from 6.4 to 9.393 benchmarks from 9.3 to 12.155 benchmarks from 12.1 to 14.925 benchmarks from 14.9 to 17.713 benchmarks from 17.7 to 20.55 benchmarks from 20.5 to 23.37 benchmarks from 23.3 to 26.27 benchmarks from 26.2 to 29.01 benchmark from 29.0 to 31.83 benchmarks from 31.8 to 34.66 benchmarks from 34.6 to 37.43 benchmarks from 37.4 to 40.21 benchmark from 40.2 to 43.11 benchmark from 45.9 to 48.71 benchmark from 65.6 to 68.4

122nd strongest of 779

Signal-to-noise +12.27

Difficulty
64thpercentile
1 benchmark from 5.9 to 14.81 benchmark from 23.6 to 32.52 benchmarks from 32.5 to 41.31 benchmark from 41.3 to 50.22 benchmarks from 50.2 to 59.08 benchmarks from 59.0 to 67.97 benchmarks from 67.9 to 76.819 benchmarks from 76.8 to 85.623 benchmarks from 85.6 to 94.541 benchmarks from 94.5 to 103.331 benchmarks from 103.3 to 112.239 benchmarks from 112.2 to 121.148 benchmarks from 121.1 to 129.973 benchmarks from 129.9 to 138.859 benchmarks from 138.8 to 147.6118 benchmarks from 147.6 to 156.5127 benchmarks from 156.5 to 165.3113 benchmarks from 165.3 to 174.235 benchmarks from 174.2 to 183.117 benchmarks from 183.1 to 191.99 benchmarks from 191.9 to 200.81 benchmark from 200.8 to 209.64 benchmarks from 209.6 to 218.5

280th hardest of 779

Capability needed ≲ 158.2

12 dated models · 4 frontier modelsClosed weightsOpen weights
Data: unLAPatlas.village.dev
#ModelSettingScoreSource
1Claude Opus 5.5max · Mini-SWE-Agent
76.4
MaxxBench leaderboard: Claude Opus 5.5 (max)
2GPT-6 Astrax-high · Mini-SWE-Agent
71.9
MaxxBench leaderboard: GPT-6 Astra (xhigh)
3GPT-5.6 Solmax · Mini-SWE-Agent
69.1
MaxxBench leaderboard: GPT-5.6 Sol (max)
4Claude Sonnet 5.5max · Mini-SWE-Agent
66.9
MaxxBench leaderboard: Claude Sonnet 5.5 (max)
5Claude Fable 5x-high · Mini-SWE-Agent
63.5
MaxxBench leaderboard: Claude Fable 5 (xhigh)
6Claude Opus 5max · Mini-SWE-Agent
63.5
MaxxBench leaderboard: Claude Opus 5 (max)
7GLM-5.3max · Mini-SWE-Agent
56.7
MaxxBench leaderboard: GLM-5.3 (max)
8GLM-5.3 Flashmax · Mini-SWE-Agent
55.1
MaxxBench leaderboard: GLM-5.3 Flash (max)
9Gemini 3.8 Flashhigh · Mini-SWE-Agent
53.4
MaxxBench leaderboard: Gemini 3.8 Flash (high)
10GPT-6 Lunamax · Mini-SWE-Agent
51.7
MaxxBench leaderboard: GPT-6 Luna (max)
11GPT-5.6 Lunamax · Mini-SWE-Agent
46.1
MaxxBench leaderboard: GPT-5.6 Luna (max)
12Qwen3.8-Flash-Nextx-high · Mini-SWE-Agent
46.1
MaxxBench leaderboard: Qwen3.8 Flash Next (xhigh)
Loading Atlas data…