GitHub Copilot

Last verified: Jun 8, 2026

41 sourced changes·98% official·verified against official docs, changelogs & Wayback·last checked Jun 8, 2026

01 / VERDICT

Caution

Frequent recent tightening — high constraint volatility right now.

WHY CAUTION — trend: unstable · drift -44% over 90d · 7 tightenings in 90d

Constraint Stability

41/100

higher = fewer recent restrictions

unstable-44% / 90d

GitHub Copilot has shown frequent recent tightening, mainly in model availability.

Most noticeable: Full removal of the GitHub Models service.

Recency

days since it last tightened

Velocity

9.1

tightenings per year

Silent

of tightenings were unannounced

Reversals

60%

of tightenings later reversed (same dimension)

GitHub Copilot constraint metrics — computed from 10 tracked tightenings · median 7 days between tightenings. Original to this site; not published by the vendor.

IMPACT BY PERSONA

Solo / occasional

Brief, occasional use is rarely touched by limit changes.

low

Power user / daily heavy

Exposed to rate / frequency tightening.

high

Team / shared workflow

Exposed to model-availability, usage-based cost changes across the team.

high

Startup / cost-scaling

Exposed to pricing volatility and usage-based cost as usage scales.

medium

Enterprise / long-term dependency

Continuity exposure: pricing-predictability, model-continuity, ownership / policy shifts.

high

02 / ANALYSIS

No tool in this set has tightened harder. Copilot has marched from flat subscriptions toward fully metered billing, stripping the guardrails as it went. It started in June 2025 with enforcement of premium-request caps (Pro at 300/mo) and overage billing. December 2025 removed the $0 default budget that had blocked all overage, letting charges accrue automatically. April 2026 pulled Opus models from Pro entirely (Pro+ only) and paused sign-ups. The decisive break came in June 2026: a switch to usage-based AI Credits, with Pro's $10 buying 1,000 credits, the fallback model removed, and roughly 4.7 million subscribers migrated — power users reported bills jumping ten- to fifty-fold, some from around $29 to roughly $750/month. Model churn runs heavy alongside the billing shift: GPT-5.2 and its Codex variant deprecated, Opus 4.6 (fast) scheduled for removal, Claude Fable 5 added then suspended days later by the same government export directive that hit Anthropic directly, and the GitHub Models service closed to new customers ahead of full retirement on July 30, 2026. There are counterweights — newer models added (Sonnet 5, Kimi K2.7, Opus 4.8), BYOK support, and crucially spend-control tooling in per-user credit budgets and hard AI-credit session limits. So the platform made metering the default and then handed back governance controls to contain the fallout. The user report says it plainly — getting less for the same price — and the practical mandate is to cap agentic sessions, watch credit consumption, and never pin to a single model.

03 / CURRENT LIMITS

Current state + recent changes

Free: Limited completions + 50 premium requests/mo
Pro ($10/mo): Usage-based AI Credits — $10 = 1,000 credits; agentic sessions ~30-40 credits; no fallback model
Pro+ ($39/mo): $39 = ~7,000 credits; keeps Opus access

Last verified: Jun 8, 2026

RECENT LIMIT & PRICING CHANGES

BeforeAfter

Previously, you could only manage this through the...→You can now manage a cost center’s AI credit pool directly in the billing UI where you create and edit cost centers.

Jul 2026·source

N/A↑You can now retrieve every user’s progress against a multi-user budget from a single REST API endpoint. This makes it much faster to find who is close to their limit

Jul 2026·source

N/A→You can now set AI credit session limits in Copilot CLI and the GitHub Copilot SDK to cap the amount an agent spends in a session.

Jul 2026·source

N/A→Per-user AI credit budgets available for cost centers

Jun 2026·source

04 / WHERE IT MATTERS

Which usage patterns the recent changes affect — each line backed by the events driving it.

Specific-model dependencehigh exposure

Jul 2026↑Expands access to Kimi K2.7 model to Business and Enterprise tiers.·source

Jul 2026↓Full removal of the GitHub Models service.·source

Jun 2026↑Users gain access to a newer version of the Claude Opus model.·source

Jun 2026↑Adds a new model option for developers within Copilot.·source

Jun 2026↓Users relying on Opus 4.6 (fast) must migrate to alternative models before the deprecation date.·source

Jun 2026↓Product being sunset: new customers blocked from onboarding; existing users keep access until full retirement.·source

Jun 2026↓Users who adopted Claude Fable 5 in Copilot lost access with no notice and must switch models. Reversal of the model addition made days earlier.·source

Jun 2026↑New model (Claude Fable 5) added across all plans; no existing model removed.·source

Jun 2026↓Users can no longer access GPT-5.2 or GPT-5.2-Codex models in Copilot Chat, inline edits, or agent modes; enterprise admins must enable replacement models via Copilot policies.·source

Apr 2026↓Opus models pulled from Pro tier·source

High-frequency usagehigh exposure

Jun 2026↓Switched to metered AI Credits, no fallback·source

Jun 2026↓Agentic usage bills jump 10-50x·source

Jun 2025↓Premium-request caps + overage billing enforced·source

Agentic / high-volume costmedium exposure

Jul 2026→Improved visibility and management of AI credit allocation for cost centers.·source

Jul 2026↑Improved visibility into budget consumption via API, reducing manual overhead for tracking user limits.·source

Jul 2026↑Adds a new model option for agentic workflows in JetBrains IDEs.·source

Jul 2026→Allows administrators to enforce hard spending caps on agentic sessions.·source

Jun 2026→New opt-in admin capability: enterprise admins can set a per-user AI credit budget on a cost center. Cost-governance tooling, not a platform-wide vendor limit.·source

Jun 2026↑Allows users to bypass standard model provider constraints by using their own API keys.·source

Jun 2026→The API now provides visibility into AI credit consumption per user, which may assist in monitoring usage-based billing.·source

Jun 2026↑Adds opt-in automation capability for reasoning-based tasks; no existing capability removed.·source

Jun 2026↓Switched to metered AI Credits, no fallback·source

Jun 2026↓Agentic usage bills jump 10-50x·source

Dec 2025↓Default $0 spend cap removed·source

Jun 2025↓Premium-request caps + overage billing enforced·source

05 / WHAT CHANGED

Technical 8 · Product 27 · Business 2 · Policy 4

Apr 2026
↓ Opus models pulled from Pro tier
productImpact Medium · source
Dec 2025
↓ Default $0 spend cap removed
policyImpact Medium · source
Jun 2025
↓ Premium-request caps + overage billing enforced
technicalImpact Medium · source

▸ View full constraint history (22 events)

Jul 2026
→ Improved visibility and management of AI credit allocation for cost centers.
businessImpact Low · source · secondary source
Jul 2026
↑ Improved visibility into budget consumption via API, reducing manual overhead for tracking user limits.
technicalImpact Low · source · secondary source
Jul 2026
→ A preview billing app is being retired and its functionality consolidated into the main billing UI with greater visibility; no loss of capability.
productImpact Low · source · secondary source
Jul 2026
↓ Full removal of the GitHub Models service.
productImpact High · source
↔ shipped alongside Enables multimodal input capabilities for Copilot chat. (Jul 2026) — more for more, not a pure cut.
Jul 2026
→ Allows administrators to enforce hard spending caps on agentic sessions.
technicalImpact Medium · source
Jun 2026
↑ Users gain access to a newer version of the Claude Opus model.
productImpact Low · source · secondary source
Jun 2026
→ New opt-in admin capability: enterprise admins can set a per-user AI credit budget on a cost center. Cost-governance tooling, not a platform-wide vendor limit.
technicalImpact Low · source · secondary source
Jun 2026
↑ Removal of a hard block on new user registrations, transitioning to a gradual rollout.
technicalImpact Low · source · secondary source
Jun 2026
↓ Users relying on Opus 4.6 (fast) must migrate to alternative models before the deprecation date.
productImpact Medium · source
↔ shipped alongside Provides a new dedicated desktop interface for Copilot. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↓ Product being sunset: new customers blocked from onboarding; existing users keep access until full retirement.
productImpact High · source · secondary source
↔ shipped alongside Improved visibility into active user counts for enterprise reporting. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↓ Users who adopted Claude Fable 5 in Copilot lost access with no notice and must switch models. Reversal of the model addition made days earlier.
productImpact High · source · secondary source
↔ shipped alongside Adds opt-in automation capability for reasoning-based tasks; no existing capability removed. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↑ Removes character constraints on repository-level custom instructions for Copilot code review.
technicalImpact Low · source · secondary source
Jun 2026
→ Adds opt-in admin control to exclude content from Copilot code review; no existing capability removed.
policyImpact Low · source · secondary source
Jun 2026
↑ Enterprises can now organize billing and usage tracking across more business units.
technicalImpact Low · source · secondary source
Jun 2026
→ Admins can IP-restrict EMU account access (web, Git, API, tokens)
policyImpact Low · source
Jun 2026
↓ Users can no longer access GPT-5.2 or GPT-5.2-Codex models in Copilot Chat, inline edits, or agent modes; enterprise admins must enable replacement models via Copilot policies.
productImpact High · source · secondary source
↔ shipped alongside New model (Claude Fable 5) added across all plans; no existing model removed. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↓ Switched to metered AI Credits, no fallback
businessImpact High · source
↔ shipped alongside New model (Claude Fable 5) added across all plans; no existing model removed. (Jun 2026) — more for more, not a pure cut.
Jun 2026
↓ Agentic usage bills jump 10-50x
technicalImpact High · source
↔ shipped alongside New model (Claude Fable 5) added across all plans; no existing model removed. (Jun 2026) — more for more, not a pure cut.
Apr 2026
↓ Opus models pulled from Pro tier
productImpact Medium · source
Dec 2025
↓ Default $0 spend cap removed
policyImpact Medium · source
Sep 2025
→ Usage analytics dashboard added
policyImpact Low · source
Jun 2025
↓ Premium-request caps + overage billing enforced
technicalImpact Medium · source

▸ Other product updates (19) — feature launches, not constraint changes

Jul 2026
↑ Adds custom instructions and configuration capabilities to Copilot code review, reducing setup friction.
productImpact Low · source · secondary source
Jul 2026
↑ Adds support for Xcode 27 on GitHub-hosted runners, expanding build environment capabilities.
productImpact Low · source · secondary source
Jul 2026
↑ Expands the scope of secret scanning coverage for users.
productImpact Low · source · secondary source
Jul 2026
↑ Increases flexibility for users to bring their own keys and models to JetBrains IDEs.
productImpact Low · source · secondary source
Jul 2026
↑ Adds new security analysis capability to the Copilot app interface.
productImpact Low · source · secondary source
Jul 2026
↑ Expands access to Kimi K2.7 model to Business and Enterprise tiers.
productImpact Low · source · secondary source
Jul 2026
↑ Adds a new model option for agentic workflows in JetBrains IDEs.
productImpact Low · source · secondary source
Jul 2026
↑ Adds a new model option to the Copilot model picker.
productImpact Low · source
Jul 2026
↑ Enables multimodal input capabilities for Copilot chat.
productImpact Low · source
Jun 2026
↑ Adds a new model option for developers within Copilot.
productImpact Low · source · secondary source
Jun 2026
↑ Allows users to bypass standard model provider constraints by using their own API keys.
productImpact Low · source · secondary source
Jun 2026
↑ Adds Claude as an agent provider in JetBrains IDEs, expanding model choice for users.
productImpact Low · source · secondary source
Jun 2026
↑ Removal of preview status indicates general availability and increased stability for CLI usage.
productImpact Low · source · secondary source
Jun 2026
→ The API now provides visibility into AI credit consumption per user, which may assist in monitoring usage-based billing.
productImpact Low · source · secondary source
Jun 2026
↑ Simplifies model selection for users across all plans.
productImpact Low · source
Jun 2026
↑ Provides a new dedicated desktop interface for Copilot.
productImpact Low · source
Jun 2026
→ Improved visibility into active user counts for enterprise reporting.
productImpact Low · source · secondary source
Jun 2026
↑ Adds opt-in automation capability for reasoning-based tasks; no existing capability removed.
productImpact Low · source · secondary source
Jun 2026
↑ New model (Claude Fable 5) added across all plans; no existing model removed.
productImpact Low · source · secondary source