other
12.15B parameters
Commercial OK
Reviewed July 2026

Mellum2 12B-A2.5B

Mellum2 12B-A2.5B is JetBrains' from-scratch SWE-focused MoE (12.15B total, 2.5B active), released June 2026 under Apache-2.0. Purpose-built for code completion and agentic SWE tasks rather than general chat; the Q4_K_M GGUF build is ~8GB.

License: Apache-2.0·Context: 131,072 tokens

Overview

Mellum2 12B-A2.5B is JetBrains' from-scratch SWE-focused MoE (12.15B total, 2.5B active), released June 2026 under Apache-2.0. Purpose-built for code completion and agentic SWE tasks rather than general chat; the Q4_K_M GGUF build is ~8GB.

Strengths

  • Apache-2.0 license with commercial use allowed
  • MoE efficiency (2.5B active of 12.15B total) at a small footprint
  • Purpose-trained on software-engineering tasks, not general chat

Weaknesses

  • Narrow training focus — general chat/knowledge quality unverified
  • New vendor entrant to open weights; little independent track record
  • GGUF ships from a community-mirror HF repo, not a first-party Ollama library tag
BLK · QUALITY BENCHMARKreviewed · raw logs

Reviewed quality benchmarks

First-party rows were run by RunLocalAI; reviewed community rows are labeled in the data. Every row links to the raw test-run log.

BenchmarkQuantRuntime / HardwareScoreRaw log
HumanEval+
tested 2026-07-20
Q4_K_M
ollama-0.32.1-vast5090
rtx-5090
76.8/100
Gist →

Q4_K_M note:First-party HumanEval+ on a rented Vast.ai RTX 5090 32GB instance via Ollama 0.32.1. Generation via evalplus_openai_generate.py, scoring via evalplus.evaluate (native Linux, no Windows shim needed). No DB credentials were placed on the rented host — results were pulled back and ingested from the operator's machine.

Want to verify? Every row links to its Gist with full stdout and stderr of the run. The runner script is in the public repo (scripts/run-humaneval-plus.ts) — reproducible end-to-end. Browse all coding scores at /benchmarks/coding.

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M8.0 GB10 GB

Get the model

Ollama

One-line install

ollama run hf.co/JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_MRead our Ollama review →

HuggingFace

Original weights

huggingface.co/JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Mellum2 12B-A2.5B.

Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Frequently asked

What's the minimum VRAM to run Mellum2 12B-A2.5B?

10GB of VRAM is enough to run Mellum2 12B-A2.5B at the Q4_K_M quantization (file size 8.0 GB). Higher-quality quantizations need more.

Can I use Mellum2 12B-A2.5B commercially?

Yes — Mellum2 12B-A2.5B ships under the Apache-2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of Mellum2 12B-A2.5B?

Mellum2 12B-A2.5B supports a context window of 131,072 tokens (about 131K).

How do I install Mellum2 12B-A2.5B with Ollama?

Run `ollama pull hf.co/JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M` to download, then `ollama run hf.co/JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M` to start a chat session. The default quantization is Q4_K_M.

Source: huggingface.co/JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.

Related — keep moving

Before you buy

Verify Mellum2 12B-A2.5B runs on your specific hardware before committing money.