Gemini 4 Hydrogen lands in Copilot as the successor to Gemini 3.8 Flash

GitHub, a Microsoft company, says Gemini 4 Hydrogen, Google’s latest Flash model, is now generally available in GitHub Copilot. It is billed at provider list pricing under usage-based billing, and the rollout is gradual.

The model is aimed at fast, efficient agentic coding, tool use and multistep workflows. GitHub says it follows instructions better than Gemini 3.8 Flash.

Which plans and tools get Hydrogen first

According to GitHub, Gemini 4 Hydrogen is available to Copilot Pro, Pro+, Max, Business and Enterprise users. Free-tier users are not listed in the changelog.

It can be chosen in the model picker in Visual Studio Code, Visual Studio, Copilot CLI, the Copilot coding agent, the Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode and Eclipse, and it is also reachable through the Copilot SDK. GitHub warns that some accounts may not see it yet and should check back.

On Business and Enterprise plans, administrators control access through the model policy in Copilot settings. Under default model enablement, new models switch on automatically unless an administrator has turned off the global default or disabled this specific model.

The changelog does not give a per-token price or a usage multiplier. It points readers to GitHub’s models and pricing documentation, so cost for a given plan has to be checked there.

A replacement with a deadline attached

The timing is tied to another change. In a separate changelog entry, GitHub announced that it will remove six models from Copilot on December 14, according to our report on that deprecation. Gemini 3.8 Flash is on that list, and Gemini 4 Hydrogen is the named replacement. Teams that have pinned Gemini 3.8 Flash in agent configurations or policies have about a month to switch.

For Business and Enterprise administrators, the practical point is the policy setting. If the global default was turned off, Hydrogen will not appear on its own and has to be enabled before the cutoff.

Cheap per task, but the evidence is still thin

This section looks at cost and economics, where Hydrogen’s pitch is clearest. Copilot bills it at list price, so the provider’s own rates apply. According to our earlier report, Google lists Hydrogen at $1.50 per million input tokens and $7.50 per million output tokens, flat across a 1M-token context window. The same report cites Artificial Analysis listing Gemini 4 Argon at $4 and $20, which makes Hydrogen 37.5% of Argon’s price on both counts.

The coding results come from a vendor. In Cognition’s Devin test, Hydrogen scored 38.2% on FrontierCode 2.0 at a mean cost of $2.40 a task, against 40.0% and $4.04 for Gemini 4 Carbon, roughly 40% cheaper. Cognition ran those tests itself, and they have not been independently reproduced.

Those numbers describe Devin, not Copilot. Copilot wraps models in its own prompts, tools and context handling, so the same model can behave differently there. GitHub’s changelog offers no benchmark of its own, only the qualitative claim of better instruction following over Gemini 3.8 Flash.

What the announcement leaves out

GitHub does not say whether Hydrogen is the default model anywhere, what its usage multiplier is for plans with included requests, or which regions get it first. Nor does it say how long the gradual rollout will take. Free-plan access is not mentioned either.

For developers choosing a model for agent work, the sensible test is a side-by-side run on their own repositories against the model they use today. Under usage-based billing, a cheaper model that needs more retries can still cost more per finished task, a point none of the sources here settle.