GitHub announced on November 13, 2026 that it will retire six model options from Copilot on December 14, affecting users and software teams relying on those choices. The notice covers Chat, inline edits, ask and agent modes, and code completions.
There is an important exception: Claude Haiku 4.5 will stay available to individual Copilot subscribers on annual plans. Those users do not need to replace it under this notice.
Which models are changing
GitHub names five suggested alternatives; Kimi K3’s replacement remains unannounced.
Retiring option | GitHub’s suggested alternative |
|---|---|
Claude Fable 5 | Claude Fable 5.2 |
Claude Haiku 4.5 (annual-plan exception above) | Claude Haiku 5.5 |
Gemini 3.8 Flash | Gemini 4 Hydrogen |
Gemini 4 Argon | Gemini 4 Carbon |
Grok 4.6 | Grok 4.7 |
Kimi K3 | Not yet announced |
Readers moving from Flash can use RohitAI’s Hydrogen access and billing guide for the successor’s rollout and account requirements.
Check access as well as saved selections
A suggested successor is not necessarily available to every account. Access depends on the plan, client and workplace policy; Free and Student offer automatic model selection only.
For Business and Enterprise, enabling access is separate from choosing a model for a session. GitHub’s default-availability rules exclude Fable 5.2, which is disabled by default. Administrators should check its explicit model policy rather than assume the general automatic-enablement rule covers it.
Individual Pro and Pro+ subscribers who kept legacy annual request-based billing after June 1, 2026 do not receive new models or features. Check that arrangement before planning a switch; the Haiku exception preserves an existing choice for annual subscribers.
For workflows that select an affected model, review the Chat picker, command-line options and saved CLI settings, and model fields in custom-agent profiles. Changing Chat’s model does not change inline suggestions. The retirement notice does not promise automatic replacement of every saved selection.
Replacement rates are not all equivalent
GitHub’s November 13 rate card lists usage-based prices in US dollars per million tokens, the units of model input and output. Fable 5.2 and Haiku 5.5 have lower listed rates than their predecessors; Grok 4.7 matches Grok 4.6 within the same context tier.
Hydrogen’s input/output rates are $1.50/$7.50, twice Flash’s promotional $0.75/$3.75. Cached input, or reused context, remains $0.075 for both. This is a unit-rate comparison, not a forecast that each task will cost twice as much.
Argon and Carbon currently list $2 input and $10 output, but both promotions end December 1. Carbon then becomes $4 input, $0.20 cached input and $20 output. Argon’s later rates are unspecified. For December budgeting, use Carbon’s December tariff, not today’s promotion.
Flash’s pricing footnote runs through December 31; that does not extend its availability beyond the December 14 retirement date.
These are usage rates, not new subscription fees. A task can involve several model calls, and consuming included credits is not itself an extra invoice charge. Paid-plan code completions and next-edit suggestions remain outside AI-credit billing.