Claude, the essentials — edition of September 9, 2026
Fable 5.1 and Mythos 5.1 Launch as Claude Code Fixes a Self-Introduced Gateway Bug
Anthropic added two 1M-context models to its lineup and pushed three rapid Claude Code point releases, including a same-week fix for a gateway sign-in regression the company had itself introduced. Separately, the status page logged two resolved error incidents spanning both new and established models.
- Claude Code shipped versions 2.1.261, 2.1.265, and 2.1.266 in quick succession, the last explicitly fixing a 2.1.265 regression that made the undocumented CLAUDE_CODE_USE_GATEWAY variable force Cloud-gateway sign-in in proxy and LLM-gateway setups.
- Anthropic launched Claude Fable 5.1 as the successor to Fable 5, plus Claude Mythos 5.1, which the release notes restrict to Project Glasswing participants; both carry a 1 million token context window.
- Prompt-cache reads on the two new models are priced at 0.025x base input cost, a steeper cache-read discount than the 0.1x applied to other Claude models.
- Two separate status-page incidents — elevated errors across Mythos 5.1, Fable 5.1, and Opus 5, and a second one on Sonnet 5 — were both marked resolved.
- The ant CLI's new `ant apply` command lets teams manage agents, environments, skills, memory stores, and deployments as version-locked code.
Claude Code patches a bug it introduced days earlier
Anthropic released three Claude Code point versions in close order — 2.1.261, 2.1.265, and 2.1.266 — most of them aimed at gateway, proxy, and enterprise-policy setups. The most notable change was corrective: version 2.1.265 added telemetry (user.email, user.groups) for Claude Desktop and Cowork sessions routed through a Claude apps gateway, but in doing so also caused the previously inert CLAUDE_CODE_USE_GATEWAY environment variable — which had done nothing unless ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKEN were also both set — to start forcing Cloud-gateway sign-in on its own, breaking existing LLM-gateway and proxy configurations. Anthropic shipped 2.1.266 specifically to reverse that regression.
The same 2.1.265 release also added folder-based plugin loading via --plugin-dir and a 1GB cap on tool results saved to disk, with truncation flagged in the preview, while the earlier 2.1.261 added a diagnostic line explaining failed organization-policy loads in /status and claude doctor, raised inline output limits with new bashOutputMaxChars and taskOutputMaxChars settings (up to 128K characters), and let large subagent system prompts be passed via file instead of the command line. Taken together, the sequence — and the fact that a fix was needed within days — shows Anthropic actively tuning Claude Code for enterprise deployment friction rather than treating it as settled, with the tradeoff that an undocumented variable's behavior shifted across point releases without a major version bump.
Sources: Claude Code 2.1.266 — Fixed a 2.1.265 regression affecting LLM-gateway and proxy setups — Claude Code · Claude Code 2.1.265 — Added user.email and user.groups to the telemetry Claude Desktop and Cowork send through a Claude apps gateway — Claude Code · Claude Code 2.1.261 — Added an "Organization policy" line to /status and claude doctor — Claude Code
Two new models: Fable 5.1 and a restricted Mythos 5.1
Per Anthropic's own release notes, Claude Fable 5.1 (claude-fable-5-1) launched as the successor to Claude Fable 5, aimed at long-running agentic coding, knowledge work, and research. Alongside it, Anthropic introduced Claude Mythos 5.1 (claude-mythos-5-1), which the release notes state is available specifically to participants in "Project Glasswing" rather than the general API. Both models support a 1 million token context window. Anthropic also disclosed a pricing detail unique to the pair: prompt-cache reads cost $0.25 per million tokens, equal to 0.025x the base input price, compared with the standard 0.1x cache-read rate applied elsewhere in its lineup, while cache-write pricing is unchanged.
Anthropic paired the launch with a guide, "Building Claude Commerce Agents," describing how to build commerce-oriented agents on Claude. The combination of a gated model (Mythos 5.1 limited to a named program) and a use-case-specific application guide published in the same window suggests Anthropic is continuing to pair new model capacity with narrower rollout and applied guidance rather than a single uniform general-availability release.
Sources: We've launched Claude Fable 5.1 (claude-fable-5-1)... alongside Claude Mythos 5.1 (claude-mythos-5-1) for Project Glasswing participants — Anthropic (release notes) · Prompt cache reads on Claude Fable 5.1 and Claude Mythos 5.1 cost $0.25 USD per million tokens — Anthropic (release notes) · Introducing Claude Fable 5.1 and Claude Mythos 5.1 — Anthropic · Building Claude Commerce Agents — Anthropic
Two error incidents, both resolved
Anthropic's status page recorded two distinct incidents the same day, both marked resolved. The first covered elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 together; Anthropic stated it had identified the cause and was working on a fix. The second, tracked separately, covered elevated errors specific to Claude Sonnet 5, for which Anthropic reported it was investigating.
The first incident grouped the two just-launched 5.1 models with the established Opus 5 in a single fault window, meaning newly released capacity and an existing production model were affected together rather than the issue being confined to the new releases alone.
Sources: Claude — Elevated errors for multiple models — status.claude.com · Claude — Elevated errors for Claude Sonnet 5 — status.claude.com
Agent platform tooling and applied notes
Anthropic also extended its infrastructure-as-code approach to agent management: version 1.30.0 of the ant CLI added `ant apply`, which creates and updates agents, environments, skills, memory stores, and deployments from files describing each resource in a repository. Users run `ant apply`, review the plan it prints, and approve it, with changes tracked through a committed claude-lock.json lockfile — a plan-review-apply pattern familiar from conventional infrastructure tooling, now applied to fleets of Claude-based agents and skills.
Alongside this, Anthropic published a getting-started guide titled "Getting Started with Claude Design" and an account of formalizing a proof of Fermat's Last Theorem, broadening the applied and research-adjacent material it released in this period.
Sources: Version 1.30.0 of the ant CLI adds ant apply — Anthropic (release notes) · Getting Started with Claude Design — Anthropic · Formalizing Fermat's Last Theorem — Anthropic
This edition is an original synthesis written by Claude from aggregated news — Anthropic's own sources first (release notes, status, newsroom, research, engineering), then the press, Hacker News, Reddit and GitHub, under the editorial supervision of Héra SASU. Every fact links to its article, publisher named. See the live feed →
Claude News is published by Héra SASU. Independent media, not affiliated with Anthropic.