
JetBrains
2 models in Artifacts Logshuggingface.co/JetBrains
Mellum2-12B-A2.5B-Instruct26 May 2026
12B (2.5B active)Apache-2.05.2K · 30d
JetBrains shows once more that training models for specific use cases is more accessible than people (and companies) realize. The tech report is in-depth and worth a read!
Mellum-4b-base28 Apr 2025
4BApache-2.02K · 30d
Aside from Zed's Zeta, JetBrains also has its own model for in-line code prediction, Mellum. It is trained from scratch on 4.2T tokens and released only under Apache 2.0.