Claude, the essentials — edition of September 5, 2026
New Frontier Models Arrive Alongside a Flurry of Claude Code Fixes and Outages
Anthropic rounded out the week by shipping two new Claude models and an infrastructure-as-code tool for agents, while Claude Code absorbed four rapid point releases and the status page logged two resolved error spikes.
- Claude Fable 5.1 and Claude Mythos 5.1 launched with 1M-token context windows; Mythos 5.1 is limited to Project Glasswing participants.
- Prompt-cache reads on both new models are priced at 0.025x base input cost, well below the 0.1x multiplier used on other Claude models.
- The ant CLI's new `ant apply` command lets teams define agents, environments, skills, memory stores, and deployments as files, reviewed via a plan and a `claude-lock.json` lockfile before changes are applied.
- Claude Code moved through four versions (2.1.258–2.1.261), fixing a macOS 12 launch regression and a remote-session content bug while adding a diff panel, higher output-capture limits, and clearer diagnostics for cache misses and blocked organization policies.
- Anthropic's status page recorded two resolved incidents: elevated errors on Claude Mythos 5.1, Fable 5.1, and Opus 5, and a separate elevated-error episode on Claude Sonnet 5.
Two New Models, One Gated
Anthropic introduced Claude Fable 5.1 as the successor to Fable 5 for long-running agentic coding, knowledge work, and research, releasing it alongside Claude Mythos 5.1. Both carry a 1 million token context window, but Anthropic specifies that Mythos 5.1 is available only to participants in Project Glasswing, making it a restricted-access sibling rather than a general-availability release.
The launch came with a pricing change specific to these two models: prompt-cache reads cost $0.25 per million tokens, equivalent to 0.025x the base input price, versus the 0.1x multiplier that applies to cache reads on other Claude models. Cache write pricing is unchanged. For workloads built around the new 1M-token window, where repeatedly re-processing long contexts would otherwise be costly, this makes heavy cache reuse markedly cheaper than on prior-generation models.
Sources: Claude Fable 5.1 and Claude Mythos 5.1 launch — Anthropic (release notes) · Prompt cache pricing for Fable 5.1 and Mythos 5.1 — Anthropic (release notes)
Treating Agents Like Infrastructure
Version 1.30.0 of the ant CLI added `ant apply`, which reads resource definitions from files in a repository and creates or updates agents, environments, skills, memory stores, and deployments accordingly. The workflow mirrors infrastructure-as-code tooling: each resource is described declaratively, `ant apply` computes a plan, and the user approves it before changes take effect, with a `claude-lock.json` lockfile meant to be committed alongside the definitions.
This positions agent configuration as something to be versioned and reviewed like code, rather than clicked together through a console, which matters for teams running several agents or environments where consistency and auditability of changes become harder to maintain by hand.
Sources: ant CLI 1.30.0 adds ant apply — Anthropic (release notes)
Claude Code's Steady Cadence
Four Claude Code releases landed in quick succession. On the fix side, 2.1.258 addressed a regression introduced in 2.1.255 that had broken launches on macOS 12 (Monterey), and separately fixed remote and scheduled sessions failing with a "non-empty content" error after a re-sent permission approval could not be applied. On the feature side, 2.1.259 added a managed setting letting organizations push HTTP/SSE MCP servers to every user, a `--permission-prompts none` flag for unattended headless hosts that auto-denies anything requiring a prompt, and recognition of GitLab merge-request commands in the tool summary and footer.
Later releases focused on visibility: 2.1.260 added a diff panel next to the conversation in fullscreen mode (toggled with `/diff`) and surfaced the likely cause of prompt-cache misses in `/cost` and the status line; 2.1.261 added an "Organization policy" line to `/status` and `claude doctor` explaining why a policy failed to load, raised inline command and background-task output limits to up to 128K characters via new settings, and let large subagent system prompts be read from a file instead of the command line. Taken together, the run of releases reads as consolidation — fixing platform regressions and giving both individual users and organizations better diagnostics — rather than a single headline feature.
Sources: Claude Code 2.1.261 release notes — Claude Code · Claude Code 2.1.260 release notes — Claude Code · Claude Code 2.1.259 release notes — Claude Code · Claude Code 2.1.258 release notes — Claude Code
Two Brief, Resolved Outages
Anthropic's status page logged two separate incidents, both since marked resolved. The first covered elevated errors on requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5, with Anthropic stating it had identified the cause and was working on a fix; the underlying cause itself was not detailed on the page. The second, affecting Claude Sonnet 5, was reported while still under investigation, with Anthropic saying it would provide an update as errors were traced.
Notably, the first incident touched Fable 5.1 and Mythos 5.1 within days of their launch, alongside Opus 5, though the status page does not attribute the errors to the new releases specifically. Both episodes were resolved without further detail published on duration or root cause.
Sources: Claude — Elevated errors for multiple models — status.claude.com · Claude — Elevated errors for Claude Sonnet 5 — status.claude.com
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.