Introducing Mellum2: A 12B Mixture-of-Experts Model by JetBrains
• 34
AI & ML interests
Reliable and context-aware coding assistance
Recent Activity
Papers
Mellum2 Technical Report
The Best of N Worlds: Aligning Reinforcement Learning with Best-of-N Sampling via max@k Optimisation
Articles
Mellum2 GGUF weights
-
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-BF16
Text Generation • 12B • Updated • 258 • 6 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q8_0
Text Generation • 12B • Updated • 1.13k • 19 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q6_K
Text Generation • 12B • Updated • 2.12k • 12 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M
Text Generation • 12B • Updated • 2.8k • 35
Series of code models by JetBrains
Mellum2 model weights
-
JetBrains/Mellum2-12B-A2.5B-Thinking
Text Generation • 12B • Updated • 5.7k • 324 -
JetBrains/Mellum2-12B-A2.5B-Instruct
Text Generation • 12B • Updated • 4.33k • 82 -
JetBrains/Mellum2-12B-A2.5B-Thinking-SFT
Text Generation • 12B • Updated • 220 • 25 -
JetBrains/Mellum2-12B-A2.5B-Instruct-SFT
Text Generation • 12B • Updated • 124 • 14
A collection of datasets, fine-tuned models and benchmarks to train your models for perfect Kotlin code generation.
Mellum2 GGUF weights
-
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-BF16
Text Generation • 12B • Updated • 258 • 6 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q8_0
Text Generation • 12B • Updated • 1.13k • 19 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q6_K
Text Generation • 12B • Updated • 2.12k • 12 -
JetBrains/Mellum2-12B-A2.5B-Thinking-GGUF-Q4_K_M
Text Generation • 12B • Updated • 2.8k • 35
Mellum2 model weights
-
JetBrains/Mellum2-12B-A2.5B-Thinking
Text Generation • 12B • Updated • 5.7k • 324 -
JetBrains/Mellum2-12B-A2.5B-Instruct
Text Generation • 12B • Updated • 4.33k • 82 -
JetBrains/Mellum2-12B-A2.5B-Thinking-SFT
Text Generation • 12B • Updated • 220 • 25 -
JetBrains/Mellum2-12B-A2.5B-Instruct-SFT
Text Generation • 12B • Updated • 124 • 14
Series of code models by JetBrains
A collection of datasets, fine-tuned models and benchmarks to train your models for perfect Kotlin code generation.