共有:

GitHub Copilot / Model Picker

Copilot adds two more models

Until last month the picker only covered a handful of brains, so matching a model to the task was hit or miss. Claude Fable 5.1 landed on Sept 1 and Gemini 3.8 Flash on Sept 3 — two flagship models, two days apart.

AI Navigate Editorial2026.09.205 min read

MODEL PICKER — SEP 2026 Sep 1 Claude Fable 5.1 Sep 3 Gemini 3.8 Flash Oct 2 older models retire Pro / Pro+ / Business / Enterprise admins can enable/disable via policy
01
Why Now

Why two flagships landed two days apart

Rival tools' model lineups had grown wide — Copilot's picker just caught up in a hurry.

GitHub Copilot's model picker got its official changelog entry on September 1, 2026 announcing general availability of Claude Fable 5.1, Anthropic's newest "Mythos"-class model built for long-horizon, autonomous coding work. Just two days later, on September 3, Google's Gemini 3.8 Flash also went GA in Copilot.

Per GitHub's own early testing, Gemini 3.8 Flash performed strongly on complex terminal-based coding tasks, with rigorous validation and persistent recovery from failures. Shipping two major models within the same week suggests Copilot's picker had started to look thin next to rivals like Cursor and Windsurf, which already offer a wide model lineup.

02
By The Numbers

This week's changes, by date

Sep 1
Claude Fable 5.1 goes GA
Sep 3
Gemini 3.8 Flash goes GA
Oct 2
older models scheduled to retire

Availability differs between the two. Fable 5.1 reaches Pro+, Business, and Enterprise plans, and on Business/Enterprise it stays off until an admin enables the policy (default is off). It also turns on data retention by default — Anthropic retains prompts and outputs to run safety classifiers, though not for model training. Gemini 3.8 Flash, by contrast, reaches the wider Pro, Pro+, Business, and Enterprise range, billed at introductory pricing through December 31, 2026.

Claude Fable 5.1Gemini 3.8 Flash
Anthropic / Mythos classGoogle / Flash class
Built for long, autonomous coding runsStrong at complex terminal tasks, validation & recovery
Pro+ / Business / EnterprisePro / Pro+ / Business / Enterprise
Admin policy off by defaultEnabled by default unless an admin disables it
03
Who Benefits

Who it helps, and how

Model-per-task switchers

Long, autonomous refactors now have a clear match in Fable 5.1, while terminal-heavy debugging has a clear match in Gemini 3.8 Flash — the split just got sharper.

PMs and admins

Rolling out Fable 5.1 on Business/Enterprise takes an explicit policy toggle — plan for it as an active setup step, not something users just discover on their own.


If you always reach for the same model, this barely registers.


04
What's Next

Watch the October 2 retirement date

This addition isn't standalone. The September 3 changelog also flags that a set of older models leave Copilot's surfaces on October 2, 2026: Gemini 3.5 Flash and Gemini 3.6 Flash both move to Gemini 3.8 Flash, Kimi K2.7 Code moves to Kimi K3, and Claude Opus 4.7 moves to Claude Opus 5. So this week's two additions are as much a forced migration as an expansion of choice.

Concretely: (1) Business/Enterprise admins should evaluate and enable the Fable 5.1 policy now rather than later, (2) teams with workflows hard-pinned to Gemini 3.5/3.6 Flash, Kimi K2.7 Code, or Claude Opus 4.7 should verify their successor model works before Oct 2, and (3) individual developers who like switching models per task should try Fable 5.1 and Gemini 3.8 Flash on real work to learn where each one is strongest. If you always reach for the same model, the only thing that actually forces your hand is the October 2 retirement.