The main reason to drop a cloud coding assistant is not the bill but the fact that your source code goes into it. Open coding models install on your own server and plug into the editor: code stays in house, there are no request limits, and the model can be fine-tuned on your repository so it knows your internal libraries and style. The range is wide: from small completion models that fit on a modest GPU to agentic ones that edit files in the repository and expect serious hardware. On hard architectural work an open model is weaker than a top cloud assistant, and responses take longer — local completion has to be tuned for latency. Review does not go away either: assistant-written code goes through the same review and tests whatever produced it.
A small coding assistant from MTS AI that understands requests in Russian. Runs locally, with plugins for VS Code and JetBrains.
The broadest open coding family: from 0.5B for autocompletion to 480B for agents. Qwen3-Coder-Next (80B, 3B active) works as a developer agent on a single GPU.
JetBrains models for fast code autocompletion. Mellum2 (12B, 2.5B active) is already a full assistant: it writes and edits code, calls tools and reasons.
Mistral models for agentic development: they read the repository, edit files and run commands on their own. The 24B version fits on a single GPU.
DeepSeek's coding model family: from small autocompletion models to the large MoE V2, which matched closed models in 2024. Later, coding moved into DeepSeek's general models.
Kodify-Nano, Qwen Coder and Mellum are Apache 2.0; the larger Devstral 2 123B uses a modified MIT license with a revenue threshold, and DeepSeek-Coder runs under the DeepSeek License with restrictions on certain uses. Read those terms before a company-wide pilot.
An open model can run on your own server: data stays in-house, there is no per-request fee, and the model can be fine-tuned on your documents.
