// RELEASE HISTORY
MAJORFEATUREPATCH
2026-05
MAJORLATEST
Claude Opus 4.8 + Mythos Preview
Released May 28, 2026. Most capable generally available Claude. Same post announces Claude Mythos — a new tier above Opus via Project Glasswing. Pricing: $5/$25 per MTok.
▾
// FULL RELEASE NOTES
›Released: May 28, 2026 — source: anthropic.com/news/claude-opus-4-8
›Model ID: claude-opus-4-8 — available via API, Claude.ai Pro, Team, and Enterprise
›Improved reasoning and coding benchmarks over Opus 4.7; sharper judgment on agentic tasks
›Fast mode: 2.5× speed at 3× lower cost than previous fast mode
›Dynamic Workflows in Claude Code (research preview): plan and run hundreds of parallel subagents in one session
›Standard pricing: $5 / MTok input, $25 / MTok output
›— WHAT'S NEXT (from same post) —
›Claude Mythos: "a new class of model with even higher intelligence than Opus"
›Project Glasswing: small number of cybersecurity orgs using Mythos Preview now
›Cyber safeguards required before GA — Anthropic confirms "swift progress"
›Mythos-class GA expected "in the coming weeks" per the Opus 4.8 post
›Source: anthropic.com/news/claude-opus-4-8
2026-05
MAJORSpaceX Compute Deal + Higher Usage Limits
May 6, 2026. Anthropic announces a compute partnership with SpaceX alongside significantly higher usage limits for Claude Pro and API users.
▾
2026-04
MAJORClaude Opus 4.7
Released April 16, 2026. Improved agentic reliability, substantially better vision with greater image resolution, and more creative professional output.
▾
2026-02
MAJORClaude Sonnet 4.6
Released February 17, 2026. Current production-tier model. Default in Claude Code CLI and SDK. Best balance of speed, cost, and capability in the Claude 4 family.
▾
2025-05
MAJORClaude 4 Family + Haiku 4.5 Launch
Full Claude 4 model family ships — Haiku 4.5, Sonnet 4.x, and Opus 4.x. All models include native tool use, extended context, and improved instruction following.
▾
2025-02
MAJORClaude 3.7 Sonnet + Extended Thinking
First hybrid reasoning model in the Claude line. Toggle standard or extended think-time per request. Visible step-by-step reasoning for hard math, coding, and logic problems.
▾
// STAY CURRENT
Full release notes and technical docs on the official site.
⚠ AI-CURATED — data may be inaccurate or outdated. Verify at official sources.
// ALWAYSBERUSHING · AI INTELLIGENCE FEED