GitHub Copilot Model Guide

A compact reference for choosing models in Copilot Chat and agent workflows. Use it to balance model quality, token cost, latency, and the type of work you are asking Copilot to do.

Compare Models Use Recommended Flows Explore Workflows

Give each model the right job.

A starting strategy, not a rule: spend more where judgment matters and less where the steps are clear.

Plan

Use GPT-6.1 Sol or another stronger model to clarify goals, compare approaches, and write testable steps.

Deliverable: a bounded plan + acceptance checks.

Build

Try GPT-6 Luna for focused edits, documentation, and implementation against that plan.

Deliverable: small changes + passing checks.

Review

Use a stronger model for risky decisions or unresolved failures. Verify results with tests and human judgment.

Escalate when the same failure repeats.

What should I hand to the cheaper model?

Give it the goal, relevant files, the agreed approach, constraints, and commands or checks that prove success. Work in small steps. Ask it to stop and report a blocker instead of guessing.

“Implement step 2 of this plan. Keep the public API unchanged. Run the listed checks. If the assumptions fail, explain what you found before changing the approach.”

Switch models in the Copilot model picker. Preserve the plan and relevant context when starting a new conversation; switching alone does not guarantee an effective handoff.

Start with the task.

Choose a situation to see a practical starting point.

What are you doing?

These are editorial recommendations, not model benchmarks. Judge the actual result, not the model’s confidence.

Same tokens. Different price.

20,000 uncached input tokens + 4,000 output tokens, at standard rates.

Prices checked October 10, 2026

Linear scale, $0 to $0.400. Gemini 3.8 Flash uses its promotional rate through December 31, 2026. This compares prices for equal tokens, not quality or cost per successful task. Input tokens are what you send; output tokens are what the model generates, including billed reasoning.

GitHub’s official rates · AI-credit billing: 1 credit = $0.01. Subscription fees, included allowances, caching, and long-context tiers are excluded here. Availability depends on your plan, client, and organization policy. Legacy annual request-based plans differ.

Make the tradeoff visible.

Adjust the work. See what model switching could save.

Estimated usage cost

Sol for every step$0.480

Sol plan + Luna build$0.100

79%less estimated usage cost

Illustrative estimate, not a benchmark. Both routes include the same planning + review steps. Every step uses the token counts you enter; real steps vary.

See the calculation

Cost = (input tokens × input rate + output tokens × output rate) ÷ 1,000,000. No cache reads or writes. Standard tier only; per-step input is capped at 272K. Dollars show usage value, not necessarily an extra charge on your bill.

How thinking effort affects usage

“High” is a setting. Success is the goal.

Start with Luna on a bounded task. If your client offers thinking effort, try a higher setting when deeper reasoning is useful. There is no published universal conversion from “High” to a fixed cost.

Check the result. If it repeats a failed approach, misses constraints, or needs architectural judgment, pause and switch. A cheap attempt that never finishes is not an efficient workflow.

Enterprise Routing Graphic

Six-Model Copilot Enterprise Route and Migration Map

A current October 10, 2026 view of five selected Copilot Enterprise routes plus GPT-5 mini migration guidance. Prices are GitHub Copilot AI-credit rates per 1M tokens; 1 AI credit equals $0.01.

Deprecation scheduled October 19 GPT-5 mini

Do not start new workflows on GPT-5 mini. Migrate fixed collection and routine prompts to GPT-5.6 Luna before October 19, 2026.

Input: $0.25
Cached / write: $0.025 / none
Output: $2.00

Cheap iteration GPT-6 Luna

Use for quick checks and lightweight coding loops; it is the lowest-cost GPT-6 route. Keep GPT-5 mini migrations on GitHub's listed GPT-5.6 Luna target until separately validated.

Input: $0.10
Cached / write: $0.01 / $0.125
Output: $0.50

Balanced coding GPT-5.6 Terra

Use for balanced everyday interactive and agentic coding. Keep tests, source checks, and human review as independent validation gates.

Input: $2.00
Cached / write: $0.20 / $2.50
Output: $12.00

Everyday development Claude Sonnet 5.5

Use for well-scoped everyday development such as feature work and bug fixes; GitHub's early testing matched Sonnet 5 coding quality with fewer steps, tokens, and tool calls plus faster completion.

Input: $2.00
Cached / write: $0.10 / $2.50
Output: $10.00

Balanced all-rounder GPT-6.1 Sol

Use for agentic coding and terminal workflows that benefit from strong multistep performance and efficient token use.

Input: $2.00
Cached / write: $0.10 / $2.50
Output: $10.00

Exceptional escalation Claude Opus 5.5

Reserve for long-running, high-consequence agent work that needs efficient multistep recovery after cheaper routes fail.

Input: $4.00
Cached / write: $0.20 / $5.00
Output: $20.00

Knowledge-base rule: self-reported confidence is not enough. Require source links, citation checks, and evidence validation before publishing a model-generated answer to a knowledge base.

Reference Table

Selected Copilot Enterprise Token Pricing

Copilot Enterprise seats contribute a pooled monthly AI-credit allowance. Usage is token-based by model and token type, one AI credit equals $0.01, and usage beyond the included pool is billed in AI Credits at GitHub's listed per-token rates.

Model	Provider	Status	Input / 1M	Cached / Write	Output / 1M	Use It For
GPT-5 mini	OpenAI	GA · deprecation scheduled Oct 19	$0.25	$0.025 cached; no cache write	$2.00	Currently GA; scheduled for deprecation on October 19, 2026. Migrate routine lightweight workflows to GPT-5.6 Luna.
GPT-5.6 Luna	OpenAI	GA	$0.20	$0.02 cached; $0.25 write	$1.20	Cheap iteration, quick checks, and lightweight coding loops.
GPT-6 Luna	OpenAI	GA	$0.10	$0.01 cached; $0.125 write	$0.50	Lowest-cost GPT-6 route for smaller, faster tasks; gradual rollout and client or policy availability can vary.
GPT-5.3-Codex	OpenAI	GA · base + LTS	$1.75	$0.175 cached; no cache write	$14.00	Agentic software development; the Business and Enterprise base model when no other model is enabled, with LTS availability through February 4, 2027.
GPT-5.6 Terra	OpenAI	GA	$2.00	$0.20 cached; $2.50 write	$12.00	Balanced everyday interactive and agentic coding; use tests, source checks, and human review for validation.
GPT-5.6 Sol	OpenAI	GA	$4.00	$0.40 cached; $5.00 write	$20.00	Long multipart orchestration and diagnosis when the higher full-rate cost is justified.
GPT-6.1 Sol	OpenAI	GA	$2.00	$0.10 cached; $2.50 write	$10.00	Agentic coding and terminal workflows with strong multistep performance and efficient token use; gradual rollout and client or policy availability can vary.
GPT-6 Astra	OpenAI	GA	$10.00	$1.00 cached; $12.50 write	$50.00	Premium OpenAI route for long-horizon autonomous coding when planning and validation quality outweigh cost.
Claude Haiku 5.5	Anthropic	GA	$0.10	$0.01 cached; $0.125 write	$0.50	Fast, high-volume work such as subagents, quick edits, and terminal tasks; gradual rollout and plan or policy availability can vary.
Claude Sonnet 5.5	Anthropic	GA	$2.00	$0.10 cached; $2.50 write	$10.00	Well-scoped everyday feature work and bug fixes with efficient task completion; gradual rollout and plan or policy availability can vary.
Claude Opus 5	Anthropic	GA	$5.00	$0.50 cached; $6.25 write	$25.00	Exceptional unresolved or high-consequence escalation after cheaper models fail.
Claude Opus 5.5	Anthropic	GA	$4.00	$0.20 cached; $5.00 write	$20.00	Long-running agentic coding and knowledge work with efficient multistep recovery; gradual rollout and plan or policy availability can vary.
Claude Fable 5.1	Anthropic	GA	$10.00	$0.25 cached; $12.50 write	$50.00	Specialized premium route only after retention review and administrator enablement; Enterprise ZDR requires approved access.
Gemini 3.8 Flash	Google	GA promo	$0.75 promo	$0.075 cached; no cache write	$3.75 promo	Cost-aware Google route for complex terminal coding tasks during introductory pricing through December 31, 2026.

Enterprise billing caveat: Auto can route around busy or degraded models, and Copilot code review automatically selects an undisclosed model. Models also vary by client, enterprise policy settings, preview access, context size, and reasoning level. Rates shown are default-tier rates; long-context usage can cost more. Legacy annual-plan multiplier language for some individual plans is not the Enterprise billing model; Selected-model usage is metered by input, cached input, cache write where applicable, and output tokens; Copilot code review also consumes GitHub Actions minutes.

Auto-selection option: GitHub offers Efficiency, Balance, and Intelligence tiers in VS Code, Copilot CLI, and the Copilot app. Efficiency favors low cost, Balance weighs cost, quality, and speed, and Intelligence favors quality for complex work. All tiers use the same eligible model pool; Auto evaluates each prompt, so even Intelligence can select a small model for a simple task. Usage follows the selected model's token rate regardless of tier, with a 10% Auto discount for paid plans. Use Auto when policy allows it, but select a model explicitly when retention, capability, or predictable per-token cost matters.

October 2 migration: GitHub deprecated Claude Opus 4.7, Gemini 3.5 Flash, Gemini 3.6 Flash, and Kimi K2.7 Code across Copilot on October 2, 2026. Move explicit selections to Claude Opus 5.5, Gemini 3.8 Flash, or Kimi K3 as applicable, and verify current availability in the model picker and organization policy.

October 19 migration: GitHub has scheduled GPT-5 mini, GPT-5.4 mini, GPT-5.4, GPT-5.5, Gemini 3.7 Flash, and Grok 4.5 for deprecation across Copilot. Move explicit selections to GitHub's listed replacements before the deadline and confirm those models are enabled by Enterprise policy.

Fable governance caveat: Claude Fable 5 and 5.1 retain prompts and outputs by default and are excluded from default model enablement. Enterprise zero-data-retention access requires GitHub approval and configuration under the current time-limited exception through the end of 2026.

HydraFusion research-preview caveat: Copilot CLI, VS Code 1.140 or later, and the GitHub Copilot app can use HydraFusion to choose among single, cascade, and independent-critique patterns from a model pool spanning multiple providers. Business and Enterprise administrators must allow preview features. Usage is billed from every model leg at its standard token rate, so treat benchmark savings as directional, inspect actual cost and quality on your own tasks, and keep normal validation and policy gates.

Task Guide

Quick Picks by Task Type

These are site recommendations grounded in GitHub's published capabilities and pricing and OpenAI's model-selection guidance. This site takes a conservative approach for new or high-risk workflows: establish an accuracy baseline with the most capable model, then compare identical inputs and test smaller models for cost and latency. For established work, keep the lightest model and settings that consistently meet the quality bar.

Small code edits

Start: GPT-6 Luna for fixed Python/API collection, known-schema edits, and cheap iteration.

Escalate: GPT-5.6 Terra when the edit needs balanced interactive coding, with independent tests and source checks.

Feature implementation

Start: GPT-5.6 Terra for balanced everyday interactive and agentic coding.

Escalate: GPT-6.1 Sol for multipart orchestration and efficient multistep validation.

Debugging production issues

Start: GPT-5.6 Terra for a balanced debugging pass, then verify logs, repro evidence, and claims independently.

Escalate: GPT-6.1 Sol for long diagnosis, then Claude Opus 5.5 only if the issue remains unresolved.

Large refactors

Start: GPT-6.1 Sol for long multipart planning and cross-file coordination when Terra is not enough.

Escalate: Claude Opus 5.5 for exceptional high-consequence review after cheaper routes fail.

Code review

Start: GPT-5.6 Terra for a focused everyday review, backed by tests, source checks, and human judgment.

Escalate: Claude Opus 5.5 for high-risk logic, security-sensitive areas, or architecture drift.

Docs and explanations

Start: GPT-6 Luna for source collection and first drafts from known evidence.

Escalate: Claude Sonnet 5.5 only when blind review shows it improves audience-fit communication.

Visual or UI reasoning

Start: Confirm visual input is supported in the active Copilot client and choose a currently supported multimodal model; do not build a new visual workflow on GPT-5 mini because it is scheduled for deprecation on October 19, 2026.

Escalate: Convert the visual evidence into explicit text requirements before moving to a deeper model whose active client supports the needed inputs.

Budget-sensitive loops

Start: GPT-6 Luna before using Terra, Sol, Sonnet 5, or Opus 5.5.

Escalate: Spend higher token-rate models only after the question, evidence, and acceptance criteria are narrowed.

Recommended Flows

How to Chain Models

Use different models for different phases instead of trying to make one model do every job.

Complex feature flow

Collect with GPT-6 Luna: gather fixed Python/API evidence and known-schema inputs.
Build with GPT-5.6 Terra: use a balanced coding pass, then check citations, extraction quality, and acceptance criteria independently.
Orchestrate with GPT-6.1 Sol: coordinate long multipart implementation or diagnosis with efficient multistep validation.
Escalate only if needed: use Claude Sonnet 5.5 when blind tests show stronger everyday development, CLI, or communication results, or Claude Opus 5.5 for exceptional unresolved risk.

Bug triage flow

Summarize with GPT-6 Luna: collect symptoms, logs, repro steps, and suspected areas.
Investigate with GPT-5.6 Terra: use a balanced agentic pass, then test whether the evidence supports the likely cause.
Diagnose with GPT-6.1 Sol: handle long cross-system reasoning when cheaper routes stall.
Escalate with Claude Opus 5.5: reserve for unresolved or high-consequence failures.

Large refactor flow

Inventory with GPT-6 Luna: find call sites, dependencies, and duplicate patterns.
Build with GPT-5.6 Terra: handle balanced coding work, then validate the evidence behind each migration claim independently.
Coordinate with GPT-6.1 Sol: manage long multipart sequencing after cheaper validation gates are exhausted.
Audit with Claude Opus 5.5: use a final pass only on the risky diff and tests.

Low-cost daily flow

Collect with GPT-6 Luna: keep fixed-source work on the lowest-cost GPT-6 route.
Iterate with GPT-6 Luna: use it for cheap checks when the task is still small.
Use GPT-5.6 Terra: move to balanced everyday coding when Luna is not enough, then validate independently.
Escalate deliberately: reserve GPT-6.1 Sol, Claude Sonnet 5.5, and Claude Opus 5.5 for their defined routes.

Cost Notes

How to Think About Cost

Copilot Enterprise usage is token-based. The real cost depends on prompt length, cached context, cache writes where applicable, outputs, selected model, and agent surface.

Use lower-cost models for proven routine work

GPT-6 Luna is the lowest-cost current GPT-6 route for routine prompts. GPT-5 mini remains available today but is scheduled for deprecation on October 19, 2026; follow GitHub's listed migration target of GPT-5.6 Luna for existing workflows, then evaluate GPT-6 Luna separately before changing production behavior. GPT-4.1 is retired from selectable Copilot use, though GitHub still lists it as a background utility model; move workflows that selected it explicitly to a supported model.

Treat mid-tier rates as workhorse capacity

GPT-5.6 Terra, GPT-6.1 Sol, and Claude Sonnet 5.5 sit in the middle of this six-model route, but they serve different jobs: Terra for balanced everyday interactive and agentic coding, GPT-6.1 Sol for efficient complex agentic coding, and Sonnet 5.5 when blind tests favor its everyday development, CLI, latency, or communication fit.

Reserve premium-heavy models

GPT-6 Astra and Claude Opus 5.5 should answer narrow, high-value questions rather than carrying every iteration of a long coding session. Consider Claude Fable 5.1 only after its retention and ZDR requirements are approved.

Sources

Official References

Verified October 10, 2026. The model list, context-window controls, reasoning controls, agent-app surfaces, sandbox policy controls, security-validation defaults, model data-retention requirements, managed-settings coverage, default model enablement, enterprise team targeting, VS Code model-provider controls, code-review billing behavior, retirements, and AI-credit billing details change often, so re-check these sources before making budget or policy decisions.

Claude Haiku 5.5 in Copilot GitHub's October 7, 2026 notice for the gradual GA rollout to Pro, Pro+, Max, Business, and Enterprise across VS Code, Visual Studio, Copilot CLI, cloud agent, the Copilot app, github.com, Mobile, JetBrains, Xcode, and Eclipse, with Business or Enterprise model-policy controls.

OpenAI model catalog OpenAI's current model lineup, capabilities, context windows, and OpenAI API pricing.

OpenAI model selection OpenAI's current guidance for choosing by workflow, comparing identical inputs, and keeping the lightest model and settings that meet the quality bar.

GPT-6 Astra guidance OpenAI's current guidance for GPT-6 Astra capabilities, prompting, tools, reasoning, and migration constraints.

Supported models GitHub's current supported model list and model availability reference.

Model comparison GitHub's capability comparison and task-fit guidance across Copilot models.

Models and pricing GitHub's current AI-credit pricing table for Copilot model usage.

Model hosting Where OpenAI, Anthropic, Google, xAI, and fine-tuned GitHub models are hosted.

Model deprecation notice GitHub's June 5, 2026 notice that GPT-5.2 and GPT-5.2-Codex are deprecated across most Copilot surfaces.

October 19 Copilot model deprecations GitHub's September 18, 2026 notice that GPT-5 mini, GPT-5.4, GPT-5.4 mini, GPT-5.5, Gemini 3.7 Flash, and Grok 4.5 are scheduled for deprecation across all Copilot experiences on October 19, with listed replacements.

GPT-4.1 deprecation notice GitHub's May 7, 2026 notice that GPT-4.1 left Copilot on June 1.

Grok deprecation notice GitHub's May 15, 2026 notice that Grok Code Fast 1 is deprecated across Copilot.

Gemini 3.5 Flash GA notice GitHub's May 19, 2026 rollout note for Gemini 3.5 Flash in Copilot.

Copilot web model changes GitHub's May 20, 2026 notice that github.com chat has a smaller model menu than the broader Copilot ecosystem.

Targeted model rules GitHub's May 26, 2026 public preview for allowing specific Copilot models by organization.

Evaluation models in Auto GitHub's June 1, 2026 notice that Copilot Individual Auto can route to evaluation models while Business and Enterprise remain GA-only.

Auto cost and quality tiers GitHub's September 14, 2026 rollout of Efficiency, Balance, and Intelligence tiers, with billing based on the selected model.

Larger context and reasoning controls GitHub's June 4, 2026 notice that supported models can use larger context windows and configurable reasoning levels with higher AI-credit consumption.

Third-party agent security validation GitHub's June 9, 2026 notice that third-party coding agents receive automatic CodeQL, dependency, and secret-scanning validation.

Claude Sonnet 5 in Copilot GitHub's June 30, 2026 notice for Claude Sonnet 5 availability, supported surfaces, gradual rollout, admin policy, billing, and zero-data-retention behavior.

Claude Sonnet 5.5 in Copilot GitHub's September 28, 2026 notice for Claude Sonnet 5.5 everyday feature and bug-fix guidance, efficiency gains, supported surfaces, gradual rollout, pricing, and model-policy controls.

Claude Opus 5 in Copilot GitHub's July 24, 2026 notice for Claude Opus 5 availability, supported surfaces, gradual rollout, admin policy, provider billing, and enhanced cyber-safety safeguards.

Claude Opus 5.5 in Copilot GitHub's September 22, 2026 notice for Claude Opus 5.5 agentic coding and knowledge-work guidance, rollout surfaces, token efficiency, error recovery, pricing, and model-policy controls.

GPT-6.1 Sol in Copilot GitHub's September 29, 2026 notice for GPT-6.1 Sol agentic coding and terminal-workflow guidance, rollout surfaces, token efficiency, pricing, and model-policy controls.

GPT-6 Sol and GPT-6 Luna in Copilot GitHub's September 22, 2026 notice for the balanced GPT-6 Sol and lightweight GPT-6 Luna routes, supported plans and clients, gradual rollout, token billing, and model-policy controls.

Grok 4.5 in Copilot GitHub's July 28, 2026 notice for Grok 4.5 availability, agentic coding fit, supported surfaces, reasoning controls, and provider-list-price billing.

Grok 4.7 in Copilot GitHub's September 21, 2026 notice for Grok 4.7's gradual rollout, agentic coding and multistep-workflow fit, supported surfaces, provider-list-price billing, and Business or Enterprise model-policy controls.

Default model enablement GitHub's July 29, 2026 notice for the August 26 Business and Enterprise policy controlling default availability of unconfigured GA models.

Enterprise team model targeting GitHub's July 31, 2026 preview for assigning optional Copilot models to enterprise teams by user role or experiment group.

Gemini model deprecation GitHub's July 31, 2026 notice that Gemini 2.5 Pro and Gemini 3 Flash are deprecated across Copilot surfaces.

Gemini 3.7 Flash in Copilot GitHub's August 13, 2026 notice for Gemini 3.7 Flash rollout, supported clients, billing, and Business or Enterprise preview policy.

GPT-5.6 in Kiro OpenAI's August 24, 2026 announcement for GPT-5.6 Sol, Terra, and Luna in AWS Kiro's spec-driven coding-agent workflow.

Hugging Face incident review OpenAI's August 26, 2026 agent-safeguards review for constrained permissions, monitoring, and failure containment in agent workflows.

Copilot weekly releases GitHub's August 28, 2026 roundup for Copilot app, CLI, VS Code, Visual Studio, Slack, Teams, JetBrains, model, usage, and permission-mode workflow updates.

Copilot in VS Code August releases GitHub's August 31, 2026 roundup for VS Code agent sessions, chat review, integrated browser, and dictation workflow updates.

Claude Fable 5.1 in Copilot GitHub's September 1, 2026 notice for Claude Fable 5.1 availability, long-horizon coding fit, and default data-retention requirements.

Copilot code review approvals GitHub's September 1, 2026 notice that Copilot code review can approve pull requests while repository merge rules still apply.

GPT-6 Astra in Copilot GitHub's September 4, 2026 notice for GPT-6 Astra availability, long-horizon coding fit, rollout, supported surfaces, and model-policy controls.

Gemini 3.8 Flash in Copilot GitHub's September 3, 2026 notice for Gemini 3.8 Flash rollout, supported surfaces, introductory pricing, and Business or Enterprise model policy.

Managed default models GitHub's September 2, 2026 notice that enterprise-managed settings can set default Copilot models globally or by team.

Copilot app and CLI content exclusions GitHub's September 2, 2026 notice that the Copilot app and CLI now honor enterprise, organization, and repository content exclusions.

Copilot JetBrains sandbox controls GitHub's September 8, 2026 notice for enterprise-managed sandbox policies, policy diagnostics, global project context, and Copilot CLI links into JetBrains IDE context.

Managed agent permissions GitHub's September 9, 2026 notice for centrally managed shell, file, and network permissions in the Copilot app, Copilot CLI, and VS Code Agent Host sessions.

MAI-Code-1-Flash deprecation GitHub's September 10, 2026 notice that MAI-Code-1-Flash is deprecated across Copilot and MAI-Code-1.1-Flash is the suggested replacement.

Project HydraFusion research preview GitHub's September 4, 2026 explanation of adaptive single, cascade, and critique workflows, benchmark limits, and standard per-model token billing.

HydraFusion in VS Code and the Copilot app GitHub's September 30, 2026 expansion of HydraFusion to VS Code 1.140 or later and the Copilot app, with clearer progress reporting and preview-policy requirements.

Copilot code-review analysis updates GitHub's September 11, 2026 notice for auto-resolved review threads, shell-backed validation behind the agent firewall, and multi-agent Lite reviews.

Selected model deprecations GitHub's September 3, 2026 notice scheduling Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7 removal from Copilot on October 2.

Copilot managed settings expansion GitHub's July 27, 2026 notice that enterprise managed settings now cover the Copilot app and Copilot cloud agent.

Copilot app access policy GitHub's July 27, 2026 notice for independent enterprise and organization access controls for the Copilot app.

Kimi K2.7 Code in Copilot GitHub's July 1, 2026 notice for its first selectable open-weight Copilot model, gradual rollout, supported surfaces, billing, and admin-policy requirements.

GitHub Models retirement GitHub's July 1, 2026 notice that the separate Models playground, catalog, inference API, and BYOK endpoints retire for all customers on July 30.

Copilot CLI security review GitHub's June 10, 2026 public preview for local AI security review in Copilot CLI.

Per-user AI-credit metrics GitHub's June 19, 2026 update adds overall per-user AI-credit consumption to organization and enterprise usage reports.

Copilot app preview expansion GitHub's June 2, 2026 notice for Copilot app canvases and agentic development workspace updates.

VS Code Copilot setup Microsoft's current setup guidance, including plan availability and built-in AI feature controls.

VS Code 1.121 release notes Microsoft's stable reference for Agents window, remote agents, AHP, utility model, and terminal-tool behavior.

VS Code 1.123 release notes Microsoft's historical release reference for agent session handoff, execution subagents, cloud-task rendering, and terminal completion behavior.

VS Code 1.124 iteration notes Microsoft's historical iteration notes for Agents window multi-chat, background send, keyboard navigation, session navigation, and Copilot CLI Agent Host routing controls.

VS Code 1.127 release notes Microsoft's historical release reference for browser-driven agent testing, per-site permissions, agent-session organization, and pull-request action banners.

VS Code 1.140 release notes Microsoft's recent stable reference for the Copilot harness, HydraFusion, multi-folder sessions, remote delegation, worktree reuse, and enterprise AI controls.

VS Code 1.141 release notes Microsoft's newest stable reference for agent worktree cleanup, cross-platform terminal sandboxing, session grids, external Copilot and Codex continuation, discovered MCP servers, and unified managed settings.