GitHub's weekly Copilot changelog (Aug. 13, covering the week of Aug. 10) adds two new models, Kimi K3 and MAI-Code-1.1-Flash, across Copilot plans. VS Code 1.133 introduces per-turn model switching within a Claude session, without restarting or losing context. Agent Plugins 1.0 is now generally available across VS Code, Copilot CLI, the SDK, and the Copilot app, with unified plugin management. Copilot CLI gains a /tasks command for subagent management, queued prompts, combined headless autopilot flags, and a /rewind command to undo changes without git. JetBrains adds Copilot Memory for cross-session context and Ollama support as a bring-your-own-key provider. An industry analyst warns that per-turn model switching creates a governance gap: nothing records which model generated which line of code, complicating debugging and audits.
Questions this post answers
Can I switch AI models mid-session in VS Code Copilot without losing context?
Yes, VS Code 1.133 lets developers switch models within a Claude session on a per-turn basis, including moving between Claude BYOK and built-in Copilot models, without restarting the session or losing context. This lets a developer start a task with one model and hand off later steps to another model mid-task. Track how IDE updates like this reshape multi-model coding workflows on daily.dev.
What new AI models were added to GitHub Copilot recently?
GitHub added Kimi K3, rolling out across Copilot Pro, Pro+, Max, Business, and Enterprise plans, and MAI-Code-1.1-Flash, which includes native image understanding along with improvements to coding quality, instruction-following, tool use, and performance. Neither model replaces existing options; they sit alongside models already available in Copilot. Developers comparing coding model options can follow model rollouts like these on daily.dev.
Does GitHub Copilot track which AI model wrote which line of code?
No, Copilot does not currently record a build attribution showing which specific model generated or modified a given line of code, even as it now supports switching between multiple models mid-session. An industry analyst described this as a governance gap, noting that model choice has effectively become a runtime dependency with no field in the build record, complicating debugging and production incident analysis. Teams weighing multi-model coding governance risks can keep tabs on this gap via daily.dev.