# Anthropic ships Claude Opus 4.8 with a 1M-token context

> Anthropic released Claude Opus 4.8 on 28 May 2026, about 42 days after Opus 4.7.

*Six weeks after 4.7, a faster, longer-memory Claude.*

By The FeaturedDaily Desk · FeaturedDaily
Canonical: https://featureddaily.com/ai-models/claude-opus-4-8-brief

> **Key:** **The one-liner:** a faster, longer-memory Claude — and a tease of the bigger model that was about to steal the headlines.

**What happened.** On 28 May 2026 Anthropic released **Claude Opus 4.8** (`claude-opus-4-8`), about 42 days after 4.7, across claude.ai, the API and Claude Code.

**The specs.** A **1M-token context** and **128K max output**, plus *effort control*, a **Dynamic Workflows** research preview in Claude Code (orchestrates parallel subagents) and a cheaper **fast mode** (~2.5× quicker).

**The claims.** Anthropic's own figures: agentic coding **64.3%→69.2%**, and **~4× less likely** than 4.7 to let flaws in its own code pass unremarked. Treat as vendor figures.

> **Note:** **Why it matters:** the post teased **'Mythos-class'** models 'in the coming weeks' — which arrived days later as Fable 5 (then got pulled by a US directive). Opus 4.8 was the warm-up.

## Key takeaways

- What: Claude Opus 4.8 (model ID claude-opus-4-8), out 28 May 2026, ~42 days after 4.7.
- Specs: 1M-token context, 128K max output, on claude.ai, the API and Claude Code.
- New: effort control, a Dynamic Workflows preview (parallel subagents), a ~2.5× fast mode.
- Claims (Anthropic's own): agentic coding 64.3%→69.2%; ~4× fewer missed code flaws.

## FAQ

### What's new in Opus 4.8?
A 1M-token context, 128K output, effort control, a Dynamic Workflows preview and a ~2.5× fast mode, plus stated coding gains — per Anthropic. Released 28 May 2026.

### Is it a big upgrade over 4.7?
Incremental on benchmarks (agentic coding 64.3%→69.2%, per Anthropic), but the fast mode and long context are the practical wins, and it claims ~4× fewer missed code flaws.