JetBrains AI Assistant 2025.2: Smarter Code, Local AI Power, and Image-Aware Chat 
JetBrains’ AI Assistant has rolled out a powerful 2025.2 upgrade, delivering sharper completions, offline flexibility, and tighter IDE integration — unlocking new ways to code smarter and faster.
Smarter Code Completion
-
Expanded language coverage: Now supports SQL, YAML, JSON, Text, and Markdown in addition to core programming languages.
-
Context-aware suggestions: Enhanced via RAG-powered understanding and the in-house Mellum LLM, ensuring more relevant results.
-
Multiline completions for Java and C++ for richer local code proposals.
Local Model Integration
-
Connect offline-friendly AI models for coding without relying on cloud:
-
Qwen2.5-Coder
-
DeepSeek-Coder 1.3B
-
CodeStral
-
Open-source Mellum
-
-
Uses OpenAI-compatible servers, enabling seamless switching between cloud and local inference.
Image-Aware AI Chat
-
Attach screenshots directly in chat when using Anthropic or OpenAI models.
-
AI can analyze images, extract code, spot errors, generate missing classes, and interpret architectures without tedious manual description.
Smarter Model Selection
-
Automatic model selection for chat actions, balancing speed, accuracy, and cost.
-
Visual indicators show:
-
Cost tier
-
Reasoning capability
-
Beta/Experimental status
-
Try It Now
-
Open the JetBrains AI widget in your IDE.
-
Follow the installation guide.
-
All AI features are free — unlimited local completions and limited cloud features.
Why it matters: This update transforms JetBrains AI Assistant into a context-smart, offline-capable, multimodal coding partner — giving developers unprecedented control over AI workflows right inside their IDE.
!