GitHub Copilot: Gemini 3.7 Flash Now Available
GitHub CopilotGitHub Copilot now supports Gemini 3.7 Flash, Google's newest Flash-tier model tuned for web and app development, coding, and long-horizon agentic workflows. The model brings gains in code quality, output presentation, codebase research, and verification on complex tasks, and early third-party benchmarks show large jumps over Gemini 3.6 Flash — including a DeepSWE score climbing from 49.0% to 65.3%. It rolls out gradually to Copilot Pro, Pro+, Max, Business, and Enterprise across VS Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, the cloud agent, and the Copilot app, billed at provider list pricing under usage-based billing. Business and Enterprise administrators must enable the Gemini 3.7 Flash Preview policy before their teams can select it.
Key Takeaways
- Rapid iteration cycle, Gemini 3.7 Flash arrived just three weeks after Gemini 3.6 Flash, built through algorithmic refinements to the same reasoning foundation rather than a full retrain.
- Major coding benchmark gains, DeepSWE scores jumped from 49.0% to 65.3% and FrontierCode scores rose from 34.4% to 43.6% compared with Gemini 3.6 Flash.
- Steep introductory price cut, list pricing is set at $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026 — roughly half the prior Flash rate — before it doubles on January 1, 2027.
- Broad Copilot rollout, the model reaches the model picker across VS Code, Visual Studio, JetBrains, Xcode, Eclipse, Copilot CLI, the cloud agent, and the Copilot app for Pro, Pro+, Max, Business, and Enterprise plans.
- Enterprise gatekeeping remains, Business and Enterprise administrators must explicitly enable the Gemini 3.7 Flash Preview policy in Copilot settings before their organization's members can select the model.
- Competitive positioning, early third-party evaluations put Gemini 3.7 Flash ahead of Claude Sonnet 5 and GPT-5.6 on several coding evals, though reviewers note it can still be more cautious in its responses than some rivals.
Sources & Mentions
5 external resources covering this update
Gemini 3.7 Flash
Hacker News
Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut
VentureBeat
Gemini 3.7 Flash launches three weeks after last model, live in Spark
9to5Google
Gemini 3.7 Flash is here with better coding, reasoning, and more
Android Authority
Google AI Just Released Gemini 3.7 Flash: A Coding and Agent Model at $0.75/1M Input Tokens
MarkTechPost
A Fast Follow-Up to Gemini 3.6 Flash
GitHub has begun rolling out Gemini 3.7 Flash, Google's latest Flash-tier model, to GitHub Copilot. Notably, the model arrives just three weeks after Gemini 3.6 Flash — Google described it as a refinement of the same reasoning foundation through algorithmic improvements rather than a full retrain. According to GitHub's announcement, the model "made improvements in web and app development and agentic coding workflows" over its predecessor, with additional gains in code quality, final-output presentation, codebase research, and verification during complex, multi-step tasks.
Benchmark Gains
Independent evaluations circulating after launch point to substantial coding and agentic improvements over Gemini 3.6 Flash. On the DeepSWE benchmark, scores rose from 49.0% to 65.3%; on FrontierCode, scores rose from 34.4% to 43.6%. Reviewers also reported stronger debugging and issue-resolution performance, higher first-pass code accuracy, and more feature-complete web app output in fewer prompts. Some early testers noted the model remains more cautious or reserved in its responses compared to rivals, and that models like Claude Sonnet 5 and GPT-5.6 still lead on certain broad knowledge-work or terminal/computer-use evaluations — but on several coding-specific evals, Gemini 3.7 Flash reportedly outperformed both.
Pricing
Gemini 3.7 Flash bills under GitHub Copilot's usage-based billing model, following the provider's list pricing (rate details are published in GitHub's models and pricing documentation). Google set an introductory price of $0.75 per million input tokens and $3.75 per million output tokens, roughly half the prior Gemini 3.6 Flash rate, in effect through the end of 2026 before it doubles to $1.50 / $7.50 on January 1, 2027.
Availability
The model is rolling out gradually to Copilot Pro, Pro+, Max, Business, and Enterprise plans, selectable via the model picker in Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot cloud agent, the GitHub Copilot app, JetBrains, Xcode, and Eclipse.
Enterprise Access
Copilot Business and Copilot Enterprise administrators must enable the Gemini 3.7 Flash Preview policy in Copilot settings before anyone in their organization can select the model. Documentation on supported models and getting-started guidance is available through GitHub's Copilot documentation, with a community feedback channel on GitHub Community discussions.