{"slug": "github-copilot-retires-4-more-models-on-october-2-act-now", "title": "GitHub Copilot Retires 4 More Models on October 2 — Act Now", "summary": "GitHub announced on September 3 that four GitHub Copilot models — Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7 — will be retired on October 2, giving developers 23 days to migrate to replacements including Gemini 3.8 Flash, Kimi K3, and Claude Opus 5. The replacements offer upgrades such as Kimi K3's 1M-token context window (4x larger than K2.7 Code's 256K) and Gemini 3.8 Flash's price of $0.75 per million input tokens and $3.75 per million output, which holds until January 1, 2027. GitHub advises individual developers who pinned deprecated models to switch in the model picker, and enterprise admins to enable replacement models in Copilot policies before the deadline.", "body_md": "GitHub quietly dropped a second retirement announcement on September 3. While developers were still processing the September 1 wave that killed six models, another four are now confirmed dead on October 2: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7. You have 23 days. Here is what to do.\n\n## The Four Models Retiring October 2\n\nThe [September 3 changelog announcement](https://github.blog/changelog/2026-09-03-upcoming-deprecation-of-selected-github-copilot-models/) was easy to miss — GitHub posted it the same week as the September 1 retirements, and most coverage focused on those six. This is the one you need to act on now.\n\n| Deprecated Model | Replacement | Key Upgrade | \n|---|---|---|\n| Gemini 3.5 Flash | Gemini 3.8 Flash | Stronger agentic coding, same price until Jan 2027 | \n| Gemini 3.6 Flash | Gemini 3.8 Flash | One replacement for both Flash versions | \n| Kimi K2.7 Code | Kimi K3 | 4x larger context window (1M vs 256K tokens) | \n| Claude Opus 4.7 | Claude Opus 5 | Stronger multi-tool agentic workflows | \n\nUnlike the September 1 wave, which swept older generation models across multiple vendors, this wave targets models that were relatively recent additions to Copilot. Kimi K2.7 Code only arrived in July. Gemini 3.6 Flash launched in August. The model churn is accelerating.\n\n## Who Actually Needs to Act\n\nMost developers are fine and can stop reading here. If you use Copilot’s default model, GitHub’s automatic routing picks a current model for you — October 2 changes nothing for you in practice.\n\nThe two groups that need to act:\n\n- **Individual developers who explicitly pinned a model.** If Gemini 3.5 Flash, Kimi K2.7 Code, or Claude Opus 4.7 is selected in your Copilot Chat model picker, that selection will fail on October 2. Open the model picker now and switch.\n- **Enterprise and Business admins who locked model policies.** This is the high-risk group. If your organization’s Copilot model policy does not include Gemini 3.8 Flash, Kimi K3, and Claude Opus 5 as enabled options, your users cannot migrate to them. The deprecated model entries disappear, and users are left without a fallback.\n\nFor admins, the fix is in Organization or Enterprise Settings under Copilot → Policies. Verify the replacement models are enabled. Do this before October 2, not the morning of.\n\n## The Replacements Are Actually Better\n\nGitHub’s model retirement pace looks aggressive on the surface, but the replacements are not cosmetic upgrades.\n\n**Kimi K3** brings a 1M-token context window, four times larger than K2.7 Code’s 256K limit. For teams using Copilot’s agent mode on large codebases, that is a real operational difference. The model [became available in Copilot on August 6](https://github.blog/changelog/2026-08-06-kimi-k3-is-now-available-in-github-copilot/).\n\n**Claude Opus 5** has been [in Copilot since July 24](https://github.blog/changelog/2026-07-24-claude-opus-5-is-now-available-in-github-copilot/) and is available on Pro+, Max, Business, and Enterprise tiers. Early testing showed meaningful gains on agentic coding tasks that require coordinating multiple tools across long sessions — exactly the use case where Opus 4.7 was the preferred pick for many enterprise teams.\n\n**Gemini 3.8 Flash** consolidates both 3.5 and 3.6 Flash into a single model at the same price: $0.75 per million input tokens, $3.75 per million output. That price holds until January 1, 2027, when it doubles — a separate planning consideration if your team runs high Copilot usage volumes.\n\n## Two More Changes Hitting September 28\n\nWhile you are in configuration mode, flag two additional Copilot changes landing September 28, per the [August 28 billing and policy announcement](https://github.blog/changelog/2026-08-28-upcoming-changes-to-github-copilot-policies-and-billing/):\n\n**Code review effort defaults to Balanced.** Copilot code review has been running in Lite mode by default. On September 28, the default switches to Balanced, which runs deeper reviews, consumes more tokens, and costs more per review. On private repositories, Balanced also draws from your GitHub Actions minutes budget. Set Lite explicitly in your repository or organization settings before September 28 if you want to avoid the switch.\n\n**Copilot Chat becomes a unified experience.** Copilot Chat on github.com, GitHub Mobile, and the cloud agent will merge into one experience governed by a single policy. The significant change: chat data retention shifts from 28 days to permanent. Review whether that affects your team’s data handling policies.\n\n## Your Action Checklist\n\n**Before October 2:**\n\n1. Check the Copilot Chat model picker — if a deprecated model is pinned, switch now\n2. Admins: go to Copilot model policy settings and enable Gemini 3.8 Flash, Kimi K3, and Claude Opus 5\n3. Audit any CI/CD pipelines, agent workflows, or dotfiles that reference a specific model name\n\n**Before September 28:**\n\n1. Set code review effort to Lite explicitly if you do not want Balanced applied automatically\n2. Review the unified Copilot experience chat retention change and update internal data policies as needed\n\nThe broader lesson: GitHub is running monthly model retirement waves, and that pace is not slowing down. Teams that hardcode specific model IDs in their workflows will keep hitting this problem every four to six weeks. Referencing model capabilities rather than model names — and leaning on GitHub’s default picker for everyday tasks — is the more maintainable approach going forward.", "url": "https://wpnews.pro/news/github-copilot-retires-4-more-models-on-october-2-act-now", "canonical_source": "https://byteiota.com/github-copilot-october-2-model-deprecations/", "published_at": "2026-09-09 03:14:51+00:00", "updated_at": "2026-09-09 03:20:11.197574+00:00", "lang": "en", "topics": ["ai-products", "developer-tools", "large-language-models"], "entities": ["GitHub", "GitHub Copilot", "Gemini 3.5 Flash", "Gemini 3.6 Flash", "Kimi K2.7 Code", "Claude Opus 4.7", "Gemini 3.8 Flash", "Kimi K3"], "alternates": {"html": "https://wpnews.pro/news/github-copilot-retires-4-more-models-on-october-2-act-now", "markdown": "https://wpnews.pro/news/github-copilot-retires-4-more-models-on-october-2-act-now.md", "text": "https://wpnews.pro/news/github-copilot-retires-4-more-models-on-october-2-act-now.txt", "jsonld": "https://wpnews.pro/news/github-copilot-retires-4-more-models-on-october-2-act-now.jsonld"}}