JetBrains Mellum2.1: From 2% to 47% on SWE-Bench

JetBrains released Mellum2.1, a 12B-parameter mixture-of-experts coding model that scores 47% on SWE-bench Verified, up from 2% for the prior version, according to byteiota. JetBrains trained Mellum2.1 with reinforcement learning as its main training loop, released it under the Apache 2.0 license, and the model runs on a 12 GB GPU.

JetBrains Mellum2.1 hits 47% on SWE-bench Verified using reinforcement learning as its main training loop. Apache 2.0, 12B MoE, runs on a 12 GB GPU. Here's how to use it. The post JetBrains Mellum2.1: From 2% to 47% on SWE-Bench appeared first on byteiota .