github.blog web signal

Gemini 3.5 Flash Lands in GitHub Copilot for All Paid Tiers

google microsoft coding tools coding-tools model-integration

Key insights

  • Gemini 3.5 Flash is available across all paid Copilot tiers simultaneously, with no phased rollout by plan level.
  • The 14x premium request multiplier is explicitly flagged as tentative, signaling GitHub may reprice as usage data comes in.
  • Cache efficiency and tool-use strength position the model specifically for multi-step agentic workflows, not just autocomplete.

Why this matters

The simultaneous five-IDE launch sets a new baseline expectation for model rollout breadth, pressuring competitors like Cursor and JetBrains AI to match coverage speed. The 14x request multiplier on a model pitched as a high-volume, agentic workhorse creates a direct tension enterprises will need to budget around, and GitHub's admission that the multiplier is tentative suggests the pricing model for agentic AI tooling is still unsettled industry-wide. For technical leaders choosing or renewing Copilot plans, the real calculus is whether Flash-tier speed justifies quota burn in automated pipelines versus reserving heavier model calls for complex reasoning tasks.

Summary

GitHub Copilot now offers Gemini 3.5 Flash as a generally available model option for all paid subscribers, following a May 19 rollout that covers Pro, Pro+, Business, and Enterprise tiers. The model runs across VS Code 1.115.0+, Visual Studio, JetBrains, Xcode, and Eclipse, making it one of the broadest simultaneous IDE launches for a new Copilot model. Google's Gemini 3.5 Flash is positioned as delivering coding quality close to Pro-tier models while operating at faster inference speeds, with high cache efficiency that suits the tight iteration loops of agentic coding workflows. The catch: it carries a 14x premium request multiplier against standard quota, which GitHub flags as tentative and subject to revision. Essentially: (GitHub, Google) are pushing a speed-optimized model into enterprise dev tooling at scale. - Available on day one across five major IDEs, not staged by editor or region - 14x request multiplier means heavy agentic use will burn quota fast for Business and Enterprise accounts - Cache efficiency is a deliberate design choice for iterative, multi-step agent tasks rather than single-shot completions The broader pattern is hyperscalers using IDE integrations as the primary distribution channel for new model capabilities, effectively making the code editor the frontline of the AI platform wars.

Potential risks and opportunities

Risks

  • Enterprise Copilot accounts running automated agentic pipelines could exhaust monthly quota significantly faster than projected, triggering overage costs or workflow interruptions before GitHub revises the 14x multiplier
  • If GitHub lowers the multiplier after enterprises have already built quota-planning assumptions around 14x, it creates pricing instability that complicates multi-year Copilot contract negotiations
  • Anthropic and OpenAI could respond by accelerating their own Copilot model updates, fragmenting enterprise developer attention and increasing evaluation overhead for engineering platform teams

Opportunities

  • Cursor, Codeium, and other IDE-native AI tools can compete directly on transparent, flat-rate pricing if GitHub's multiplier model draws enterprise pushback
  • Google Cloud resellers and enterprise account teams gain a concrete upsell hook tying Gemini model access to broader Workspace or GCP commitments via the Copilot integration
  • Observability and quota-management tooling vendors (Datadog, Grafana, internal platform teams building LLM cost dashboards) face immediate demand from enterprises trying to track Copilot request burn across agentic workflows

What we don't know yet

  • Whether the 14x multiplier applies uniformly to agentic loop calls or only to certain request types within the supported IDEs
  • No benchmark comparison published by GitHub against Copilot's existing Claude 3.5 Sonnet or GPT-4o options on coding evals
  • Timeline for multiplier revision is unspecified, leaving Enterprise procurement teams without stable pricing data for Q3 2026 planning

Originally reported by github.blog

Read the original article →

Original headline: Gemini 3.5 Flash Generally Available for GitHub Copilot — Near-Pro Coding Quality at Flash Speed Across All Paid Copilot Tiers