AI Agent StoreJetBrains · Publisher checkpoint verified
Mellum2.1 Thinking
A coding model you can run on your own hardware. Use it to investigate a failing test, draft a fix and check the result while keeping repository access under your control.
Released 2026-10-08 · First reviewed 2026-10-12 · Checked 2026-10-15.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Try it on your hardware and tasks before relying on it.
What you can use today
You download the model files and supply the hardware and software to run them. Capacity is the amount of information a request can contain; settings and available memory can limit what fits.
Exact checkpoint ID: JetBrains/Mellum2.1-12B-A2.5B-Thinking. JetBrains describes a 12B model with 2.5B parameters active at each step. It supports reasoning and tool calls; your application supplies the shell, files and permissions. Its published benchmark results are vendor measurements, so they are not mixed into our independent rankings.
The publisher labels the checkpoint Apache 2.0. The card documents vLLM setup with reasoning and tool parsers. Official GGUF files are now available, although the announcement still says they are coming soon; choose a format supported by your local runner.
The model card specifies 131,072 tokens of context. Its example output budget is a request setting, not a verified maximum output limit. Allow room for the prompt, tool results and generated answer.
No verified hosted token price or measured running cost is assigned. Local use still requires hardware, memory and setup. Test changes in a disposable copy of your project before granting access to important files.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Include the time and hardware needed to run it.
Compare caching and long-context costs ↗