Nimbus models

Nimbus Labs publishes a family of models for coding on local hardware. Explore the released checkpoints and quantized files, then evaluate a model on your own project with a compatible runner.

Browse the official Nimbus local coding collection on Hugging Face

Publisher positioning: quick drafting and focused code edits.

• Model card and BF16 checkpoint
• GGUF files and release instructions
• Licenses and notices

Publisher positioning: implementation and debugging with a balance of size and capability.

• Model card and BF16 checkpoint
• GGUF files and release instructions
• Licenses and notices

Publisher positioning: more involved coding and reasoning tasks.

• Model card and BF16 checkpoint
• GGUF files and release instructions
• Licenses and notices

The Transformers releases retain multimodal components; the corresponding GGUF releases are text-only. Check the specific repository for available files and runtime instructions. Download size does not include all runtime memory needs: context cache and buffers also require capacity.

The cards credit Qwen3.5 as the foundation. Nimbus-9B v2.1 identifies DeepReinforce’s Ornith-1.0-9B as its immediate upstream. Nimbus post-training and release packaging do not replace that attribution.

The repositories label their license “nimbus-layered-open-source”. Read each release’s LICENSES.md, included license texts and third-party notices for the applicable terms. A repository label alone does not establish conformity with every definition of open-source AI.

notch provides a coding workspace and CLI; Nimbus is a model family. Configure a compatible local runner, start with a narrow change, inspect the diff and run the repository’s tests. These task descriptions are publisher positioning, not a new FreeSyntax comparative benchmark or hardware-fit guarantee.

Model selection and licensing guide · Local AI coding workflow

Model cards checked November 4, 2026. The linked publisher repositories remain the source for current files and terms.