Google shipped Gemini 3.6 Flash instead of the Pro everyone was waiting for

Google DeepMind released Gemini 3.6 Flash and 3.5 Flash-Lite on 21 July — a cheaper, faster tier built specifically for agentic coding and tool-calling workloads — while Gemini 3.5 Pro, originally due in June, is still stuck in limited preview with no confirmed date.

On 21 July, Google DeepMind released three new models — Gemini 3.6 Flash, 3.5 Flash-Lite, and a specialised 3.5 Flash Cyber build for vulnerability research — and teased Gemini 4. What it didn’t ship was Gemini 3.5 Pro, the flagship model originally slated for June that’s now missed at least two rumoured dates and remains in limited enterprise preview with no confirmed launch window.

The detail that matters more than the miss is what Google chose to ship in its place. Gemini 3.6 Flash is explicitly positioned for “agents that reason, call tools, inspect visual inputs, edit code, and continue across multiple steps” — this is a model built for the agentic coding workflows that Cursor, Claude Code, Windsurf and their peers run on, not for chat. It’s also cheaper and more token-efficient than its predecessor, cutting token usage by up to 17% at a lower price point.

This is the second time in a month a major lab has led with a cheaper, faster tier while its flagship model slips (Anthropic did something similar with Fable 5’s export-control saga). Read together, it suggests the commercial pressure in AI right now is on the cost and speed of models doing agentic, tool-calling work — the exact category that AI coding assistants and AI-native product features sit in — rather than on chasing frontier reasoning benchmarks that most production use cases don’t actually need.

So what

If you’re specifying which model powers an AI feature or an internal coding agent, the flagship isn’t automatically the right default. A cheaper, faster model tuned for tool-calling and multi-step agent work will often out-perform a more expensive general-purpose flagship on the tasks that actually make up an agentic workflow — and the cost difference compounds fast at production volume. This is a live decision on every AI product build we scope, not a one-off pick made at kickoff. If you’re weighing model choice for an AI feature or agent, see our AI products work or get in touch to talk through the tradeoffs for your build.

← All research

More research.

GitHub secret scanning cuts false alarms for AWS, Google and Alibaba cloud keys and adds a SpaceX detector

GitHub's 5 November 2026 secret scanning update tightens detectors for AWS, Google Cloud and Alibaba keys, relevant to teams building software with AI coding tools that can paste credentials into code.

Read ›

GitHub's Copilot code review can now be managed by API, with review effort levels and AI credit costs shown

GitHub's 6 November 2026 update lets organisations manage Copilot code review by API and see the review effort and AI credits each AI code review used, a sign AI-assisted development is becoming auditable.

Read ›

GitHub's Copilot impact dashboard now counts chat and VS Code agent use, and drops the 'passive users' label

On 6 November 2026 GitHub added Copilot Chat and the VS Code Agents window to its Copilot impact dashboard, bringing it to nine tracked experiences, and renamed 'Passive users' to 'Chat-first users'.

Read ›

Claude Code 2.1.319: a setting that denies permission prompts so AI agent runs don't stall

Claude Code 2.1.319 (6 November 2026) adds a permissions.denyInsteadOfPrompting setting and moves long MCP tool calls to the background, a sign AI-assisted development is being built for unattended runs.

Read ›

GitHub now reports how often developers act on Copilot code review comments

On 5 November 2026 GitHub added code review metrics, including an addressed rate, to the Copilot usage metrics API. For AI software development buyers, it is the first first-party number for how much AI review feedback is acted on.

Read ›

Dependabot can now keep AI agent skills and prompts up to date

GitHub's Dependabot gained Agent Package Manager (APM) support on 3 November 2026, treating AI agent skills, prompts and instructions as dependencies that need updating like any other software.

Read ›

Start a project

Building with AI? Let's talk.

Tell us what you are building. We'll give you an honest view of the right approach.

Start a projectSee our work