Signed up for Devin Teams yesterday.
The deal is incredible:
Unlimited SWE-2, until 10 Oct, and that model is powerful!
----------
On DeepSWE 1.1:
GPT-6 Astra — 74.1%
SWE-2 — 73.0%
GPT-5.6 Sol — 72.7%
Kimi K3 — 68.5%
Grok 4.6 — 67.5%
Fable 5.1 — 67.4%
----------
Also includes:
Fable 5.1 ✅
Opus 5.5 ✅
GPT-6 Astra
GPT-6 Sol/Luna
GLM-5.3
DeepSeek Pro / Flash models
Gemini models
Kimi K3
Grok
... and finally - Fusion. A mix of models I have yet to experience.
No idea how they did the fusion, since most of these models are not open source. I assume they somehow mix the inputs from both models and review it by a third? 🤔
@_nexuslink was super impressed just in the first 20 minutes of using Devin!