News

Gemini 4 Neon Now Generally Available in GitHub Copilot

· GitHub Blog (Changelog)

GitHub announced on November 18, 2026 that Google's Gemini 4 Neon is generally available in GitHub Copilot, aimed at agentic coding, multistep tool use, and knowledge work. Google says the model approaches Gemini 4 Carbon in quality while costing fewer tokens per task, responding sooner, making fewer unrequested edits, and avoiding unsupported claims about code, tool, or test outcomes. It is open to Copilot Pro+, Max, Business, and Enterprise users across many IDEs, the CLI, the coding agent, and mobile, billed at provider list prices. The release shows Copilot deepening its multi-model strategy.

GitHub has announced, via its changelog dated November 18, 2026, that Gemini 4 Neon from Google is now generally available inside GitHub Copilot. The release adds another third-party frontier model to Copilot's growing picker and reinforces GitHub's position as a model-agnostic layer for AI-assisted development.

What the model is designed for

According to the announcement, Gemini 4 Neon targets agentic coding, multistep tool use, and general knowledge work. Those are the workloads that increasingly define competition among coding assistants, as developers move from inline autocomplete toward autonomous agents that plan, call tools, edit multiple files, and verify their own results.

The performance claims come from Google rather than from independent testing. Google says Neon delivers quality that comes close to its larger Gemini 4 Carbon model while using fewer tokens to finish a task, which would translate into lower cost per task. Google also reports that Neon begins responding faster than Carbon, makes fewer changes that the user did not ask for during coding work, and is less prone to asserting unsupported conclusions about code behavior, tool output, or test results. These last two points address common complaints about agentic models: scope creep in diffs and overly confident reporting of success when tests have not actually passed.

Pricing and billing

The model is billed at the provider's list pricing under Copilot's usage-based billing. This indicates that premium models in Copilot are now metered according to underlying provider rates instead of being bundled at flat rates. GitHub points users to its documentation on models and pricing for specifics. For organizations, this makes the token-efficiency claim for Neon commercially meaningful, since a model that completes tasks with fewer tokens directly reduces spend.

Availability and surfaces

Gemini 4 Neon is offered to users on Copilot Pro+, Max, Business, and Enterprise plans. It can be chosen in the model picker across a broad set of surfaces: Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot coding agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse. The rollout is gradual, so some users may need to wait before the model appears.

The breadth of supported clients is notable. Having a new model reach the CLI, the cloud-based coding agent, mobile, and several competing IDE ecosystems on day one shows how much of Copilot's value now lies in distribution and integration rather than in any single model.

Administrative controls

For Copilot Business and Enterprise, administrators govern access via the model policy in Copilot settings. Under the default model enablement behavior, new models are switched on automatically unless the administrator has disabled the global default or has specifically turned this model off. Enterprises with strict governance or data-handling requirements should review these settings, since default-on enablement means new models can reach developers without a separate approval step.

Industry context and implications

The announcement sits among a cluster of related Copilot changelog entries from the same week, including the availability of OpenAI's GPT-6.2 Astra and GPT-6.2 Sol, Microsoft's MAI Code 2 and MAI Code 2 Flash, Copilot Fusion updates, and the GitHub Copilot app. Together they portray a platform that hosts models from Google, OpenAI, and Microsoft side by side, letting developers pick based on cost, latency, and behavior.

For Google, inclusion in Copilot provides exposure to a very large developer base that might otherwise default to rival models, and the emphasis on lower cost, faster first response, and restrained edits is positioned squarely against competitors in the agentic coding market. For GitHub, a steady cadence of model additions limits its dependence on any one supplier and gives enterprises leverage over cost-performance tradeoffs.

The key open question is whether Google's claims hold up in real-world developer workflows. Since the figures are vendor-reported, practitioners will likely compare Neon against Carbon and rival models on their own repositories, looking at cost per completed task, diff quality, and honesty about test outcomes. If the claims are borne out, Neon could become a default choice for high-volume agentic tasks where cost matters, while larger models remain reserved for harder problems.

Overall, this is a routine but strategically meaningful model addition: it extends Copilot's multi-vendor lineup, highlights the move toward usage-based billing, and underscores that efficiency and reliability, not just raw benchmark quality, are becoming the primary selling points for coding models.

Referenced links