GitHub Copilot to retire six models on December 14, including Fable 5 and Gemini 4 Argon

GitHub will retire six Copilot models on December 14, including Claude Fable 5 and Gemini 4 Argon. The November 13 notice applies across Copilot Chat, Ask and Agent modes, inline editing and code completions. Individual subscribers on annual Copilot plans will keep access to Claude Haiku 4.5 under an exception.

Developers who specify a model in a workflow or integration face a change before the deadline. GitHub has asked them to move to a supported model and has named five alternatives so far. GitHub will remove the retired models from Copilot itself. Which replacements show up for a developer depends on their plan, the app they use and, for work accounts, their organization's access policies.

Five replacements ahead of the deadline

GitHub points Claude Fable 5 users to Fable 5.2 and recommends Claude Haiku 5.5 for Haiku 4.5. The company suggests Gemini 4 Hydrogen for Gemini 3.8 Flash, Gemini 4 Carbon for Gemini 4 Argon and Grok 4.7 for Grok 4.6. An alternative for Moonshot AI's Kimi K3 is still to be announced.

The suggested models arrived on different dates. GitHub announced general availability for Fable 5.2 on November 9, Carbon on November 10 and Hydrogen on November 13. GitHub began rolling out Grok 4.7 on September 21 and announced general availability for Haiku 5.5 on October 7. Each launch notice describes a gradual rollout.

Two entries in December’s list were recommended during an earlier change. When support for Gemini 3.5 Flash and 3.6 Flash ended on October 2, GitHub pointed users to 3.8 Flash. It also suggested Kimi K3 for teams switching off Kimi K2.7 Code. Teams that moved to either model in October face another change if they still specify it in a workflow.

Hydrogen, Haiku 5.5 and Grok 4.7 are listed for Copilot Pro, Pro+, Max, Business and Enterprise. The individual tiers listed in the launch notices for Fable 5.2 and Carbon start at Pro+, and both models are offered to Business and Enterprise. GitHub provides model access to Free and Student users through automatic selection.

Work accounts must enable Fable 5.2

On Business and Enterprise accounts, eligible models can inherit the organization's default access policy. An administrator can turn off that default or block a model individually. The enterprise can also decide whether its organizations set their own policies. Once a model is enabled, developers can see it in the chat model picker in supported Copilot apps.

Fable 5.2 has its own access rule. GitHub's launch notice says the model is off by default for Business and Enterprise, and an administrator must enable it. Anthropic normally retains prompts and responses to run its safety classifiers. GitHub says eligible enterprises can receive zero-data-retention access through the end of 2026. The administrator still has to enable the model after the enterprise is approved.

Some alternatives change the cost per token

For usage-based Copilot plans, both the model's rate and the number of tokens processed affect the cost of a request. GitHub lists Fable 5 at $10 per million input tokens and $50 per million output tokens. Fable 5.2 is listed at $8 and $40, respectively. The promotional rates for Gemini 3.8 Flash are $0.75 for input and $3.75 for output, while GitHub lists Hydrogen at $1.50 and $7.50 on the same basis.

On December 1, GitHub's listed prices for Carbon will rise from $2 to $4 per million input tokens and from $10 to $20 per million output tokens. Code completions and suggestions for what to edit next remain unlimited on paid plans and use no AI credits. Existing annual Pro and Pro+ subscribers on request-based billing have a separate set of model multipliers.