--- title: JetBrains releases Mellum2.1, an open 12B coding model... description: JetBrains has released Mellum2.1, an updated open mixture-of-experts coding model with 12B total parameters and 2.5B active per token, licensed under Apache... canonical: https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters | daily.dev og:description: JetBrains has released Mellum2.1, an updated open mixture-of-experts coding model with 12B total parameters and 2.5B active per token, licensed under Apache... og:url: https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn og:image: https://api.daily.dev/og/posts/6P0gyF4Dn.png og:image:alt: JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters **[Collections](https://daily.dev/sources/collections)** · 2 min read · 0 upvotes · 0 comments ## Summary JetBrains has released Mellum2.1, an updated open mixture-of-experts coding model with 12B total parameters and 2.5B active per token, licensed under Apache 2.0. Unlike the architecture-level changes seen in prior releases, this update focused on reinforcement learning in sandboxed environments to improve agentic coding behaviors like codebase exploration, file editing, and self-checking. It shows benchmark gains over Mellum2, Qwen3.5-9B, and Gemma 4 E4B in agentic coding, competitive programming, math, and tool calling, and is about 1.6x faster per request thanks to multi-token prediction. It's aimed at powering self-hosted coding agents and sub-agents. Weights are available on Hugging Face now, GGUF builds are coming soon, and vLLM already supports deployment, with speculative decoding support to follow once the MTP head ships. ## Content JetBrains has released Mellum2.1, an update to its open mixture-of-experts coding model. It has 12B total parameters, of which 2.5B are active per token, and it ships under Apache 2.0. The architecture is the same as Mellum2. What changed is the training: Mellum2.1 went through extensive reinforcement learning in real sandboxed environments. The goal was to get better at the things a coding agent actually does, which is exploring a codebase, editing files, and checking its own changes. ## Benchmarks and speed Compared with Mellum2, Qwen3.5-9B and Gemma 4 E4B, Mellum2.1 shows gains in agentic coding, competitive programming, math and tool calling. Multi-token prediction makes it about 1.6x faster per request, and it is the fastest of the group under heavy load. ## Who it's for JetBrains positions it as a model for powering coding agents and sub-agents that you can host yourself. A small active-parameter count and fast inference suit that kind of work, where an agent may make many calls in a loop. ## Availability The weights are on Hugging Face now, and GGUF builds are coming soon. The vLLM project says Mellum2.1 is ready to deploy with vLLM today. Support for the upcoming MTP head will follow, which will allow faster inference through speculative decoding. ## Questions this post answers ### What is new in JetBrains Mellum2.1 compared to Mellum2? Mellum2.1 keeps the same architecture as Mellum2 but was trained differently, going through extensive reinforcement learning in real sandboxed environments to improve agentic coding tasks like exploring a codebase, editing files, and checking its own changes. It has 12B total parameters with 2.5B active per token, ships under Apache 2.0, and is about 1.6x faster per request due to multi-token prediction. _Teams evaluating self-hosted coding agent models can track model updates like this one on daily.dev._ ### How many parameters does JetBrains Mellum2.1 have and how many are active? Mellum2.1 has 12 billion total parameters as a mixture-of-experts model, with only 2.5 billion active per token. This small active-parameter count combined with fast inference makes it suited for coding agents that make many repeated calls in a loop, such as exploring a codebase or verifying edits. _Anyone sizing infrastructure for self-hosted coding agents can follow model specs like these on daily.dev._ ### Is JetBrains Mellum2.1 ready to deploy with vLLM? Yes, the vLLM project states Mellum2.1 is ready to deploy with vLLM today. GGUF builds are coming soon, and support for the upcoming MTP head will follow, enabling faster inference through speculative decoding. Weights are already available on Hugging Face under Apache 2.0. _Developers choosing inference stacks for open coding models can track deployment support on daily.dev._ ## Similar posts on daily.dev - [JetBrains open-sources Mellum2 to go where Claude Code can’t](https://daily.dev/posts/jetbrains-open-sources-mellum2-to-go-where-claude-code-can-t-hpsuvgcq0) · The New Stack · 21 upvotes · 7 comments - [Introducing Mellum2: A 12B Mixture-of-Experts Model by JetBrains](https://daily.dev/posts/introducing-mellum2-a-12b-mixture-of-experts-model-by-jetbrains-ztgyhiwra) · Hugging Face · 3 upvotes · 0 comments - [Mellum2 Goes Open Source: A Fast Model for AI Workflows](https://daily.dev/posts/mellum2-goes-open-source-a-fast-model-for-ai-workflows-njuqs9a8q) · JetBrains · 4 upvotes · 0 comments - [Agents 2.0 is here\!](https://daily.dev/posts/agents-2-0-is-here--n2hstyuvk) · Daily Dose of Data Science \| Avi Chawla \| Substack · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#jetbrains](https://daily.dev/tags/jetbrains), [#vllm](https://daily.dev/tags/vllm), [#mixture-of-experts](https://daily.dev/tags/mixture-of-experts) [View this post on daily.dev](https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters","url":"https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn"},"datePublished":"2026-10-08T14:17:46.376Z","dateModified":"2026-10-08T14:18:20.251Z","description":"JetBrains has released Mellum2.1, an updated open mixture-of-experts coding model with 12B total parameters and 2.5B active per token, licensed under Apache...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,jetbrains,vllm,mixture-of-experts","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"JetBrains releases Mellum2.1, an open 12B coding model with 2.5B active parameters"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/jetbrains-releases-mellum2-1-an-open-12b-coding-model-with-2-5b-active-parameters-6p0gyf4dn#faq","mainEntity":[{"@type":"Question","name":"What is new in JetBrains Mellum2.1 compared to Mellum2?","acceptedAnswer":{"@type":"Answer","text":"Mellum2.1 keeps the same architecture as Mellum2 but was trained differently, going through extensive reinforcement learning in real sandboxed environments to improve agentic coding tasks like exploring a codebase, editing files, and checking its own changes. It has 12B total parameters with 2.5B active per token, ships under Apache 2.0, and is about 1.6x faster per request due to multi-token prediction. Teams evaluating self-hosted coding agent models can track model updates like this one on daily.dev."}},{"@type":"Question","name":"How many parameters does JetBrains Mellum2.1 have and how many are active?","acceptedAnswer":{"@type":"Answer","text":"Mellum2.1 has 12 billion total parameters as a mixture-of-experts model, with only 2.5 billion active per token. This small active-parameter count combined with fast inference makes it suited for coding agents that make many repeated calls in a loop, such as exploring a codebase or verifying edits. Anyone sizing infrastructure for self-hosted coding agents can follow model specs like these on daily.dev."}},{"@type":"Question","name":"Is JetBrains Mellum2.1 ready to deploy with vLLM?","acceptedAnswer":{"@type":"Answer","text":"Yes, the vLLM project states Mellum2.1 is ready to deploy with vLLM today. GGUF builds are coming soon, and support for the upcoming MTP head will follow, enabling faster inference through speculative decoding. Weights are already available on Hugging Face under Apache 2.0. Developers choosing inference stacks for open coding models can track deployment support on daily.dev."}}]} ```