Mythos

TL;DR

Claude Mythos is 📝Anthropic's anticipated next-generation 📝Artificial Intelligence (AI) model. As of April 2026, no official release date has been confirmed; reporting, leaked benchmarks, and prediction-market activity point to a first-half 2026 window.

Overview

Claude Mythos is the reported working name for Anthropic's next frontier model, positioned above Claude Opus 4.6. A research preview — 📝Claude Mythos Preview — launched April 7, 2026 under Project Glasswing with limited access. General availability has not been announced. This memo tracks what is confirmed, what has been reported, and what remains speculative.

Key Facts

  • Developer: Anthropic
  • Tier: Above Claude Opus 4.6
  • Preview launch: April 7, 2026 (Project Glasswing, restricted access)
  • General availability: Not announced as of April 2026
  • Model ID (Preview): claude-mythos-preview
  • Prediction markets: Polymarket contracts active on Anthropic's next frontier model

Timeline of Key Events

  • Late 2025 – Early 2026 — References to a model codenamed or branded "Mythos" begin appearing in leaked materials and industry discussions. The name surfaces alongside descriptions of a tier above Claude Opus with substantially higher compute requirements.
  • Q1 2026 — Polymarket contracts related to Anthropic's next frontier model attract significant trading volume. Betting activity suggests market participants expect a release or announcement in the first half of 2026.
  • March 2026 — Secondary reporting increases. Multiple outlets reference "Claude Mythos" or "Anthropic Mythos" as the working name. Benchmark data — reportedly from internal evaluations — begins circulating, showing gains on SWE-bench, cybersecurity evaluations, and multi-step reasoning tasks.
  • Late March – April 2026 — Anthropic is reported to have briefed government stakeholders on the model's capabilities, particularly cybersecurity implications. This is consistent with the company's Responsible Scaling Policy.
  • April 7, 2026 — Claude Mythos Preview launches via Project Glasswing, a defensive-cybersecurity research preview restricted to participating organizations.

What Is Confirmed

  • Anthropic is developing models more capable than the current Claude Opus tier
  • The company operates under a Responsible Scaling Policy requiring safety evaluations before releasing frontier models
  • Polymarket activity and media reporting consistently reference "Mythos" as the name or tier designation
  • Claude Mythos Preview launched April 7, 2026 under Project Glasswing

What Is Not Confirmed

  • A general-availability release date
  • Whether "Mythos" is the final consumer-facing name or an internal/pre-release designation
  • Public pricing and availability across API, Claude.ai, and enterprise surfaces
  • Whether Claude Mythos will ship through AWS Bedrock, Google Cloud Vertex AI, and Microsoft Foundry at general launch

What to Watch For

  • Official Anthropic blog posts or announcements
  • Updates to the Anthropic API model list
  • Changes to Claude.ai subscription tiers
  • Responsible Scaling Policy disclosures, which typically precede or accompany major model launches
  • Transition language from "Mythos Preview" to "Mythos" or "Mythos-class" successor models

FAQ

When will Claude Mythos be released?

As of April 2026, Anthropic has not confirmed a general-availability release date. A restricted research preview — Claude Mythos Preview — launched April 7, 2026 under Project Glasswing. Prediction markets place broader availability in mid-to-late 2026.

Is Claude Mythos available to the public?

Not yet. Access is limited to Project Glasswing participants, a set of organizations focused on defensive cybersecurity. See 📝Claude Mythos Context Window for the full Preview access list.

Is "Mythos" the final name?

Unconfirmed. Anthropic has not officially branded the model. Reporting and prediction markets consistently use "Mythos," and the Preview ships under claude-mythos-preview, but the final consumer-facing name may differ.

How will we know when Claude Mythos ships?

Watch Anthropic's blog, the API model list, Claude.ai tier changes, and Responsible Scaling Policy disclosures — which generally precede or accompany frontier model launches.

Related

As builders who ship production software on 📝Claude Code and the Anthropic API daily, we track Claude model releases the way a carpenter tracks new tool catalogs. Each generation has meaningfully changed what we can build and how fast we can build it. The leaked benchmarks for Claude Mythos suggest this won't be an incremental update. If the software engineering and agentic gains are real, the calculus shifts on what kinds of systems a small team can build autonomously. We are watching closely — and building in a way that's ready to take advantage of whatever ships.

Contexts

Created with 💜 by One Inc | Copyright 2026