AI Coding News: July 7, 2026 — Fable 5 Goes Metered, Claude Code & OpenCode Ship New Builds
Claude Fable 5's included-usage window closes today, July 7, 2026 — from tomorrow, every Fable 5 token runs through metered usage credits at $10 per million input tokens and $50 per million output tokens, the pricing Anthropic set when it redeployed the model globally on July 1 after a two-week export-control suspension. The deadline lands the same day Claude Code and OpenCode both shipped fresh builds.
The Fable 5 Grace Period Ends Today
Pro, Max, and Team subscribers have been able to use Fable 5 for up to 50% of their weekly usage limit since Anthropic restored the model on July 1 following its June 12 export-control pull. That inclusion window closes July 7 — standard Enterprise seats never had it to begin with. Starting tomorrow, Fable 5 usage across Claude.ai, the Claude Platform, Claude Code, and Cowork bills through usage credits at standard API rates, though Anthropic says it intends to fold Fable back into subscription plans once capacity allows.
Claude Code v2.1.202: Bigger Workflows, Fewer Rough Edges
Claude Code's July 6 release adds a Dynamic workflow size setting in /config, letting you steer how large Claude's agent workflows run (small, medium, or large agent counts), plus new workflow.run_id and workflow.name OpenTelemetry attributes for tracing workflow-spawned agents. The rest of the release is mostly fixes: Remote Control commands failing in interactive sessions, images dropped without captions, mTLS handshake failures during certificate rotation, and /code-review reclaiming multi-agent duty now that /review <pr> reverted to a fast single pass. Full list in the Claude Code changelog.
OpenCode v1.17.14 Adds a Code Mode MCP Adapter
OpenCode's own July 6 release introduces a code mode MCP adapter for running confined orchestration scripts against connected MCP tools, hiding the raw execute tool unless code mode is explicitly enabled — a safer default for teams wiring MCP servers into automated pipelines. The release also improves GitHub Copilot model routing and ships a unified provider-connect flow plus better terminal handling in the desktop app. See the OpenCode changelog for the rest.
Also This Week: Copilot Retires Two Gemini Models
GitHub quietly flagged the upcoming deprecation of Gemini 2.5 Pro and Gemini 3 Flash in Copilot's model picker, nudging users toward Claude Sonnet 5 — now generally available in Copilot — and the newly added Kimi K2.7. It's a small note next to the Fable 5 deadline, but it's one more sign that the model roster inside every major coding assistant keeps turning over faster than most teams can requalify their defaults.
What This Means for Developers
If your team leans on Fable 5 for heavy agentic runs, check consumption before tomorrow's switch to metered credits — at $10/$50 per million tokens, an unmanaged Fable 5 habit gets expensive fast. If you're already on Claude Code, the new dynamic workflow size setting is worth tuning before spinning up large multi-agent jobs. And if your team default-routes to Gemini inside Copilot, start requalifying against Claude Sonnet 5 or Kimi K2.7 now, before the deprecation forces the switch.
References & Further Reading
Comments
Share your thoughts and join the conversation
