Datacurve AI

DeepSWE

A long-horizon software engineering benchmark from Datacurve for measuring frontier coding agents on original tasks drawn from active open-source repositories.

In index
Agentic
Source ↗

Models36

Top result75.4%Muse Spark 1.3

Top-3 spread1.2 pts

Year2026

Result and price

Results

36 results

#ModelResult

Muse Spark 1.375.4%

Claude Opus 5.574.2%

DeepSeek V4.1 Flash74.2%

GPT-6 Astra74.1%

Gemini 3.8 Flash73.8%

06COSWE-273.0%

GPT-5.6 Sol72.7%

MiMo-V2.6-Pro71.9%

Grok 4.771.0%

GPT-5.6 Terra69.6%

Claude Opus 568.8%

GPT-6 Sol68.8%

MiMo-V2.6-Flash67.9%

14STStep 5 Preview67.7%

15MAKimi K367.5%

Claude Fable 5.167.4%

GPT-5.6 Luna67.2%

GLM-5.366.9%

GPT-6 Luna66.6%

Grok 4.665.9%

Gemini 3.7 Flash65.3%

22TEHy4 preview64.3%

GLM-5.3-Flash63.4%

DeepSeek V4 Pro 081362.7%

Muse Spark 1.259.3%

Qwen3.8-Flash-Next58.7%

Qwen3.8-Omni-Flash57.8%

Qwen3.8 Max56.6%

29OAOrnith-1.5-397B56.0%

DeepSeek V4 Flash 073154.4%

Muse Spark 1.153.3%

Grok 4.553.0%

Gemini 3.6 Flash49.0%

Qwen3.8-27B42.2%

35POLaguna S 2.140.4%

36OAOrnith-1.5-35B-A3B22.0%