Summary
JetBrains/Mellum2-12B-A2.5B-Thinking is a Mixture-of-Experts code model with 12B total parameters, 2.5B active, and a 131K-token context.
It outputs explicit `<think>...</think>` reasoning traces for debugging, planning, and agentic coding. JetBrains recommends using the Instruct variant for direct low-latency answers without reasoning traces.