Skip to content

Benchmark

SWE Atlas - Test Writing

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
3
Scores recorded
4
Sources
2
IQ weight
No weight published

Score distribution

41.8%
Mean
47.8%
Median
11.5%
Std dev
30.8%
10th percentile
57.8%
90th percentile
30.8%
Min
30.8%57.8%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Hy4 previewofficial model-card appendix· Reported by Tencent Hy· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.57.8%
2LongCat 2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim47.8%
3MiniMax M3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim30.8%
4MiniMax M3Claude Code · 4C8G · 3h · four-run mean· Reported by MiniMax Research· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.30.8%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats