CloudCodeTree LogoCloudCodeTree
AI NewsTutorialsAbout
CloudCodeTree Logo
CloudCodeTree
  • AI News
  • Tutorials
  • About
← Back to AI News
Copilot: Fable 5's 50% Usage Cap Lifts July 7 — Set a Hard Spending Cap First

Copilot: Fable 5's 50% Usage Cap Lifts July 7 — Set a Hard Spending Cap First

Chris Harper

2 min read

Jul 5, 2026 · 04:14 UTC

AI
News
Developer Tools
Claude Code

Fable 5 in GitHub Copilot exits its 50% weekly usage cap on July 7 — full access restores, but it bills at Anthropic token rates and requires 30-day data retention. Set a hard spending cap before you run agents.

Fable 5 cap ends July 7. When Fable 5 was re-enabled in GitHub Copilot on July 1 (following the US export-control lift), access was capped at 50% of weekly usage limits. That cap ends July 7. Copilot Pro+, Max, Business, and Enterprise users will have uncapped access to Fable 5 from that point. Two things to know before enabling it:

  • Token-rate billing, not flat credits. Fable 5 bills at Anthropic's published rates ($10/MTok input, $50/MTok output) under Copilot's usage-based plan — not the flat premium-request credits used for Sonnet or Opus. A single long agentic session can run $50–500+ depending on context size and iteration count.
  • Data retention required. Anthropic's safety classifiers for Fable 5 require retaining prompts and outputs for up to 30 days. All other Claude models in Copilot (Opus 4.8, Sonnet 4.5, Haiku 4.5) still operate under Zero Data Retention.

Agentic billing warning. Since Copilot moved to token-based AI Credits on June 1, teams running agentic sessions on frontier models are reporting bills 10x–50x above their previous flat-rate spend. GitHub's own research found agentic tasks can consume ~1,000× more tokens than single-turn queries; at token-rate pricing, an unmonitored overnight agent run can exhaust a month's budget.

Action: Go to Settings → Billing → GitHub Copilot → spending limit and set a hard cap now — before July 7, before running any Fable 5 agent session.

Why it matters: Fable 5 is the most capable model in the Copilot picker; it's now fully unleashed. The billing model means it's also the most expensive, with no ceiling unless you set one.

Sources: GitHub Changelog: Fable 5 GA for Copilot | Fable 5 on GitHub Copilot — fable5.app | GitHub Copilot Billing Shock: Agentic Users Face 10x Cost Surge — TechTimes | GitHub Docs: usage-based billing