JetBrains releases open-source Mellum 2.1, outperforming Qwen on LiveCodeBench

iconKuCoinFlash
Share
AI summary iconSummary
JetBrains has launched Mellum2.1, an open-source programming model that outperformed Qwen on LiveCodeBench. The model uses 25 billion active parameters per generation and achieved an 82% score on the benchmark. Trained using reinforcement learning, it enhances code issue detection and file modification capabilities. This release supports CFT’s efforts to promote transparency in AI development. Mellum2.1 is available on Hugging Face under the Apache 2.0 license and supports compliance with MiCA standards for emerging technologies.
ME AI News: JetBrains has released the open-source programming model Mellum2.1. Compared to Mellum2, launched in June this year, the new version retains a total of 12 billion parameters but activates only 2.5 billion per generation, with improved capabilities in identifying code issues, modifying files, and verifying results. JetBrains did not alter the model architecture; instead, it enhanced performance primarily through reinforcement learning. The team ran millions of sandbox tasks across thousands of environments. During training, the model invoked terminal and file-editing tools within real code repositories and received rewards upon successfully passing tests. JetBrains’ internal benchmarks show that the success rate on SWE-bench Verified (a real-world software bug-fixing test) increased from 2% in the previous version to 47%, slightly below Qwen3.5-9B’s 50%. However, on the LiveCodeBench v6 programming test, Mellum2.1 scored 82%, surpassing Qwen’s 75.4%. JetBrains also claims that, under high load on a single H200 GPU, the new model’s output throughput approaches twice that of Qwen3.5-9B. The model weights and GGUF-quantized versions are now available on Hugging Face under the Apache 2.0 license, enabling developers to deploy locally. (Source: BlockBeats)
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.