Gemini Code Assist gets latest Gemini 2.5 Pro with context management and rules

· 1 min read
Gemini Code Assist

Google rolled out a batch of May upgrades to Gemini Code Assist in a post dated June 12, 2025, offering the new features immediately through its VS Code and JetBrains extensions.

The headline change is a switch to the Gemini 2.5 Pro model, which ships with a 1 million-token context window (a 2 million-token option is in testing) and now tops the LMArena leaderboard for reasoning and coding tasks.

The larger model powers fuller chat answers, longer code generation, deeper refactors, and a pull-request reviewer that scans diffs for logic gaps before they reach human teammates.

Gemini Code Assist

Developers can tailor behavior through two new levers:

  1. Custom commands that store reusable prompts in the IDE.
  2. Project-level Rules that pin the assistant to internal conventions such as mandatory unit tests or house libraries.

Context management also improves: entire folders—or the whole repo—can be injected via an @ operator, the Context Drawer shows exactly what’s in scope, and tabbed chat lets users juggle multiple problems without losing history.

Individuals still get a free tier, while organizations pick up policy controls and the upgraded review agent through Gemini for Google Cloud; GitHub connectivity is optional.

In an internal study, teams using the tool were 2.5 times more likely to finish typical coding tasks, and early community benchmarks show higher accuracy than Copilot on context-heavy queries, though responses can lag by around ten seconds.

Gemini 2.5 Pro is already live in AI Studio and the Gemini Advanced subscription, and Google says an upcoming “Deep Think” mode will extend reasoning for very large codebases and complex maths.

Source