Update Copilot model recommendations

[pimmink]

Summary

• Refresh the task-based recommendations in the Copilot model comparison page for current general-purpose, fast, deep-reasoning, and visual tasks.
• Remove outdated claims, including the unsupported lowest-cost assertion and GPT-5 mini as the default recommendation.
• Add GPT-5.4 nano to the comparison data and remove Qwen2.5, which is not present in the current Copilot release-status or plan-availability data.
• Clarify that model availability differs by plan and direct readers to the supported-models page.

Validation

• Compared the model list against GitHub Docs' model release-status and supported-plans tables.
• Confirmed model references used in the page exist in the Copilot variables file.
• Verified the official GPT-5.4 nano documentation and model capabilities.
• The Liquid table syntax was preserved; the live documentation rendering was previously confirmed to work, so no speculative bracket/Liquid changes were made.

Scope

Only the model comparison page and its data table are changed. The supported-models page already reflects the current model and plan tables, so it did not need edits.

[github-actions]

How to review these changes 👓

Thank you for your contribution. To review these changes, choose one of the following options:

• Spin up a codespace
• Set up a local development environment

A Hubber will need to deploy your changes internally to review.

Table of review links

Note: Please update the URL for your staging server or codespace.

The table shows the files in the content directory that were changed in this pull request. This helps you review your changes on a staging server. Changes to the data directory are not included in this table.

Source	Review	Production	What Changed
copilot/reference/ai-models/model-comparison.md	fpt ghec	fpt ghec

Key: fpt: Free, Pro, Team; ghec: GitHub Enterprise Cloud; ghes: GitHub Enterprise Server

🤖 This comment is automatically generated.

[Copilot]

🟡 Changes recommended

GPT-5.4 nano is recommended without disclosing its Codex-only availability and absence from Copilot Chat.

1 open finding

• Label GPT-5.4 nano as Codex-extension-only · New

What changed in this PR

Refreshes GitHub Docs guidance for choosing Copilot models by task.

Changes:

• Updates recommendations for general-purpose, fast, reasoning, and visual tasks.
• Adds GPT-5.4 nano and removes Qwen2.5 from comparison data.
• Clarifies plan availability and recommends Auto as a starting point.

File	Description
data/​tables/​copilot/​model-comparison.yml	Adds GPT-5.4 nano and removes Qwen2.5.
content/​copilot/​reference/​ai-models/​model-comparison.md	Refreshes task recommendations and availability guidance.

🧠 Review effort: Max

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🔵 Needs a closer look

A newly recommended model lacks its documented data-retention warning, and the pricing guidance points to an availability page.

0 open findings

1 resolved since last review

• Label GPT-5.4 nano as Codex-extension-only

Previously missed (2)

In code that hasn't changed since last review

Correct pricing claim and document client availability

content/​copilot/​reference/​ai-models/​model-comparison.md:27

GitHub Copilot: The supported-models page lists availability, not prices. Its pricing reference (content/copilot/reference/ai-models/supported-models.md:75) only links onward to the billing page already linked at line 25. Remove the pricing claim here. Also mention client availability, since the recommended Gemini models are unavailable on GitHub.com (data/tables/copilot/model-supported-clients.yml:125–141). Use [AUTOTITLE] as required by content/contributing/style-guide-and-content-model/style-guide.md:690.

Warn that Claude Fable 5.1 retains prompts and outputs by default

content/​copilot/​reference/​ai-models/​model-comparison.md:101

GitHub Copilot: Claude Fable 5.1 retains prompts and outputs by default, unlike other Claude models (content/copilot/reference/ai-models/model-hosting.md:59–68). That section also documents restrictions on its zero data retention exemption. Readers choosing a model for private code need this distinction before following the new recommendation. Add a brief warning and link to the retention requirements beside it.

🧠 Review effort: Max

[Copilot]

🟡 Changes recommended

The new data-retention guidance needs a factual correction before publication.

1 open finding

• Clarify data retention for Claude Fable models · New

🧠 Review effort: Max

[Copilot]

🟢 Approval recommended

Remaining feedback consists of non-blocking documentation refinements.

0 open findings

1 resolved since last review

• Clarify data retention for Claude Fable models

Previously missed (1)

In code that hasn't changed since last review

Use Liquid variables for model and Copilot names

content/​copilot/​reference/​ai-models/​model-comparison.md:123

GitHub Copilot: Use the existing Liquid variables for both MAI model mentions here and on line 127, and for "Copilot". The names are defined in data/variables/copilot.yml:251 and data/variables/product.yml:333. Hardcoding them bypasses centralized naming updates and the variable convention in content/contributing/style-guide-and-content-model/style-guide.md:883. This is non-blocking.

🧠 Review effort: Max

[subatoi]

Thanks for contributing. Since the scope of the proposed changes here is broad, and because some of it is based around personal preference and some of the changes would need to pass through an internal review process, I'm going to transfer this for an internal review to let the relevant team decide if any action is required. Sorry this won't be the exact outcome you were hoping for, but in general, if you believe you've identified any specific issues with the docs that aren't solvable with straightforward and targeted pull requests, we ask that you please open an issue for us to review before going ahead and raising a PR.