Skip to content

Benchmark

SWE-Atlas · Codebase Q&A

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

Models measured
6
Scores recorded
6
Sources
1
IQ weight
0.53

Score distribution

52.3%
Mean
53.5%
Median
8.5%
Std dev
46.2%
10th percentile
64%
90th percentile
37.9%
Min
37.9%64%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.53 for chat models, measured over 6 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

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.64%
2Muse Spark 1.3max · public QnA split · 124 tasks · mini-swe-agent · rubric grading · mean pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.59.4%
3GPT-5.6 Solmax · provider model card · public QnA split · mean pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.53.5%
4Claude Opus 5max · provider model card · public QnA split · mean pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.52.7%
5Muse Spark 1.2xhigh · public QnA split · 124 tasks · mini-swe-agent · rubric grading · mean pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.46.2%
6MiniMax M3Mini-SWE-Agent · 4C8G · 3h timeout· Reported by MiniMax Research· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.37.9%

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.