Visual Studio Code 1.126 has begun its rolling release cycle, and the first batch of changes is already live on the release notes page. While the full feature set will land throughout the month, the early signals are clear: agent UX polish and model management simplification are the headline themes.
What’s New in 1.126 So Far
Unified Model Customization Picker (June 16)
The context configuration and reasoning effort controls have been merged into a single model customization picker. Instead of adjusting these settings in two separate places, developers can now tune context windows and thinking effort from one unified interface. This is a meaningful quality-of-life improvement — especially for teams juggling multiple models with different reasoning capabilities and cost profiles.
Inline Chat Renaming (June 15)
Chat sessions in the Agents window can now be renamed directly from their tab. It’s a small UX change, but it reflects a maturing product surface. As agent sessions become a primary workflow — not an experiment — naming and organizing them becomes table stakes.
Pixel Spinner Alignment (June 18)
Chat session list items now use the same pixel spinner as the Agents window. Visual consistency across agent surfaces signals that Microsoft is treating the chat + agent experience as a unified whole, not a collection of tacked-on features.
The Bigger Picture: Where VS Code Is Headed
These early 1.126 changes sit on top of a massive wave of VS Code and GitHub Copilot investment over the past few months:
Agents Window Goes Stable
The Agents window launched in preview earlier this year and is now available in VS Code Stable. It’s an agent-first workspace surface: launch, monitor, and review agent sessions across projects without switching contexts. Remote agents over SSH/Dev Tunnels, session sync across machines, /chronicle productivity queries, and multi-session side-by-side views are all live.
BYOK & Multi-Model Management
VS Code’s model flexibility story continues to expand:
- Air-gapped BYOK — bring-your-own-key models can now run in isolated environments without GitHub auth
- Custom Endpoint providers — add any compatible chat completions or responses API
- Configurable utility models — pick which models handle tasks like commit messages, summaries, and intent detection
- Reasoning effort controls from the model picker — fine-tune the quality/latency/cost balance
Enterprise Governance Catches Up
Enterprise controls are keeping pace with the agentic explosion:
- Native MDM delivery for managed Copilot settings (Windows & macOS) — no per-user sign-in required to enforce policy
- Disable bypass permissions — admins can kill “yolo mode” (auto-approve) across the org
- Sandboxed MCP servers — local AI tools can run in restricted environments
- Command risk assessment — AI-generated safety explanations before terminal confirmations
Microsoft’s Small Model Play: MAI-Code-1-Flash
Microsoft’s MAI-Code-1-Flash small-tier coding model is rolling out in VS Code. It’s designed for utility tasks where you don’t need a big reasoning model — commit messages, quick completions, lightweight refactors. Smart token economics.
What This Means for Teams
If you’re running VS Code in an enterprise or consulting context, a few things stand out:
- Agent-first is not a trend, it’s the direction. Every release this year has deepened agent integration. If you haven’t experimented with the Agents window, remote agents, or Copilot CLI with agent mode, this is the time.
- Model flexibility is becoming a requirement. The BYOK and multi-provider investments mean VS Code is preparing for a world where teams mix models (cheap for utilities, expensive for reasoning). Start thinking about which models go where.
- Governance is catching up, but you need to configure it. The controls exist — MDM delivery, managed settings, sandboxing, risk assessment — but they don’t apply themselves. Enterprise teams should be auditing their Copilot policies against what’s now available in 1.124-1.126.
The 1.126 rolling release notes will fill up over the coming weeks. We’ll update this post as major features land. In the meantime, check the official release notes and start planning how these changes fit your workflow.
Big Hat Group helps organizations navigate the AI coding tool landscape — from VS Code and Copilot deployment to governance, security, and training. Get in touch if you need a hand.