GitHub Copilot Adds GPT-5.5 With a 7.5x Credit Multiplier, Plus JetBrains Agent Mode

GitHub Copilot shipped two notable updates on April 24. GPT-5.5 is now generally available in Copilot, and JetBrains IDE users got inline agent mode. Both are significant, and both come with caveats worth understanding before you enable them.

GPT-5.5 in Copilot: Strong Model, Expensive Credits

GPT-5.5 is rolling out to Copilot Pro+, Business, and Enterprise across VS Code, Visual Studio, JetBrains, Xcode, Eclipse, github.com, GitHub Mobile, and the Copilot CLI. GitHub calls it their strongest model yet for “complex, multi-step agentic coding tasks.”

The catch is the pricing. GPT-5.5 launches with a 7.5x premium request multiplier. If you’re on a plan with a monthly credit allotment, using GPT-5.5 burns through it seven and a half times faster than the standard rate. GitHub calls this “promotional pricing,” which implies the multiplier could change, but there’s no timeline on that.

For Business and Enterprise plans, administrators have to explicitly enable a GPT-5.5 policy in Copilot settings before users can access the model. It won’t appear in the model picker for employees unless the policy is turned on at the org level.

The rollout is gradual, so not everyone on eligible plans will see GPT-5.5 in their model picker immediately. When it does appear, it sits alongside the other available models as a user-selectable option.

OpenAI released GPT-5.5 last week for ChatGPT and Codex, where it showed meaningful gains on Terminal-Bench 2.0 (82.7%) and SWE-Bench Pro (58.6%). The GitHub Copilot integration brings those same capabilities into the IDE, though Copilot’s agent workflows differ from Codex’s headless task execution model.

The 7.5x multiplier is the real decision point for most teams. Unless your Copilot plan has headroom to absorb the credit burn, using GPT-5.5 for routine tasks will eat through your monthly allotment quickly. It makes more sense as a selective tool for genuinely hard agentic tasks than as a default model for autocomplete or chat.

JetBrains Gets Inline Agent Mode

The second update is more straightforwardly useful. GitHub Copilot for JetBrains IDEs now includes inline agent mode, accessible through Shift+Cmd+I on Mac or Shift+Ctrl+I on Windows.

Previously, Copilot’s agent capabilities in JetBrains required opening the separate chat panel. Inline agent mode brings the same functionality into the editor without the panel switch. You can invoke it in context, see results inline, and stay in the file you’re working on.

Also in this update: Next Edit Suggestions now shows inline previews before you accept, and can navigate to suggested edits in other parts of the same file using gutter indicators. A new global auto-approve setting lets you skip confirmation prompts for all tool calls, which GitHub flags with a security warning given it includes terminal commands and file edits. There are granular controls to scope auto-approve to specific tools if you want the convenience without the blanket permission.

Inline agent mode ships as a public preview, and Business/Enterprise admins need to enable “Editor preview features” in the Copilot policy settings for their users to access it.

JetBrains IDEs cover a substantial portion of the professional Java, Kotlin, and Python developer market. Copilot has had a persistent UX gap on JetBrains compared to VS Code, where the agent experience has been more polished for longer. Inline agent mode closes some of that gap.

---

Sources: GitHub Changelog: GPT-5.5 GA for Copilot, GitHub Changelog: JetBrains Inline Agent Mode