code

Mellum2.1-12B-A2.5B-Thinking

JetBrains Mellum2 line small MoE coding model with thinking mode.

JetBrainsCzechia

Figure from the Mellum2.1-12B-A2.5B-Thinking model card.

Levelmodel

Classcode

Parameters12B

Context131,000

Released8 Oct 2026

Opennessopen weights permissive

Runs onlaptop

Statusactive

From the model's own page; who made each picture and where it comes from are listed on the Image Credits page.

Benchmark Results

vendor-reported

HumanEval+coding

thinking modevendor-reported

91.5

050100

Score in %. Higher is better.

huggingface.co

LiveCodeBenchcoding

thinking mode, LiveCodeBench v6vendor-reported

050100

Score in %. Higher is better.

huggingface.co

MBPP+coding

thinking modevendor-reported

79.4

04080

Score. Higher is better.

huggingface.co

SWE-bench Procoding

thinking modevendor-reported

050100

Score in %. Higher is better.

huggingface.co

SWE-bench Verifiedcoding

thinking modevendor-reported

050100

Score in %. Higher is better.

huggingface.co

Terminal-Bench 2.1agents and tool use

thinking modevendor-reported

17.4

050100

Score in %. Higher is better.

huggingface.co

Where this comes from

Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

R&D Topics

Code language modelsproduced by Reasoning via large-scale RLproduced by

Identity

Developer: JetBrains
Family: not in the notes
Released: 8 Oct 2026 sourcesregistry dateunverified
Weights published: –
Record checked: 10 Oct 2026

Where this comes from

Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture: MoE
Total parameters: 12B sources
Active parameters: 2.5B
Checkpoint parameters: –
Context length: 131,000 tokens
Input: text
Output: text

Where this comes from

Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness: open weights permissive
Licence: Apache License 2.0 sources
Access routes: open weights download

Where this comes from

The licence file governs over registry tags and aggregator labels.

Deployment

Runs on: laptop
Footprint: T5 estimate (unverified). GGUF available. Tech report May 29.
Runtimes: not in the notes

Where this comes from

Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support: not stated
Regional languages: not in the notes
Jakarta hosting: unknown
Languages: not in the notes

Where this comes from

Indonesia fields are carried on every model; unknown is an allowed value.

Lineage

The notes record no lineage for this model.

Industry Use Cases

No industry use case links this model.

Links

Developer

JetBrainscompany, Czechia

Model

Mellum2.1-12B-A2.5B-ThinkingJetBrains, code

R&D TopicsCode Language Models Reasoning via Large-Scale RL

Use CasesChinese Medical LLMsHealthcareUse cases of models in the same R&D topic; none is recorded for this model itself.

Who Runs Itnone recordedDeployments of these use cases; none names this model.

Graph

Organisations 1

JetBrainscompany, CzechiaDeveloper

R&D topics 2

Reasoning via Large-Scale RLReinforcement learning on outcomes, often verifiable math or code, to produce long reason…produced byCode Language ModelsPretraining and fine-tuning language models on source code for completion, infilling, gen…produced by

Sources

huggingface.co/JetBrains https://huggingface.co/JetBrains
huggingface.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking https://huggingface.co/JetBrains/Mellum2.1-12B-A2.5B-Thinking

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Mellum2.1-12B-A2.5B-Thinking", https://aurion.semai.id/models/jetbrains-mellum2-1-12b-a2-5b-thinking/, records of October 2026.