DeepSWE 1.1

Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better.

Gemini 4 Argon by Google holds the best DeepSWE 1.1 result among tracked models, at 77.9% (released Sep 30 2026). Higher scores are better on this benchmark.

1. Gemini 4 Argon (Google) — 77.9%; 2. Muse Spark 1.3 (Meta) — 75.4%; 3. GPT-6.1 Sol (OpenAI) — 75.2%; 4. DeepSeek-V4.1-Flash (DeepSeek) — 74.2%; 5. GPT-6 Astra (OpenAI) — 74.1%.

32 tracked models have a published DeepSWE 1.1 score. Scores are collected from published lab reports and benchmark sources, with recorded sources credited alongside the results. We do not re-run the evaluations.

DeepSeek-V4.1-Flash by DeepSeek is the highest-ranked model with downloadable weights on DeepSWE 1.1, scoring 74.2% at rank 4 overall.