Anthropic's new version of Mythos training is complete, and Sonnet 5 is set to be released this week.

AI Startup Anthropic Completes Training of New Version Mythos Model; Claude Sonnet 5 Model Label Appears on Developer Platform; Market Anticipates Official Release Within This Week
(Background summary: Hurry to unlock Fable, Mythos! Anthropic promises closer ties to the White House "reconciliation proposal")
(Additional background: Elon Musk: Chinese large models to match Anthropic Fable by Q1 2027! Zhipu Tangjie: Won't take that long)

Table of Contents

Toggle

• Sonnet 5 Model Label Appears on Developer Platform
• Mythos Model Family: Anthropic’s “Invisible Arms Race”
• Accelerating AI Model Release Pace: From Quarterly to Weekly

The internal model competition at AI startup Anthropic is entering a heated phase. Well-known AI observer Andrew Curran revealed on X platform Monday that Anthropic has completed training of a new, more powerful Mythos model, but the final name of the new model remains unknown—possibly called Mythos 5.1, Mythos 6, or kept internal as a future development engine.

Sonnet 5 Model Label Appears on Developer Platform

Almost simultaneously, tech media Digi discovered a model label named claude-sonnet-5 on Anthropic’s partner service provider platform. Wes Roth later confirmed this discovery on X, and the developer community generally believes Claude Sonnet 5 will be officially released this week.

If the information is accurate, this will be another product line upgrade since Anthropic launched Fable 5 and Mythos 5 earlier this year. The Claude Sonnet series is positioned as a “performance and cost balance” medium-sized model, optimized for reasoning speed and API cost efficiency, making it the most widely deployed level for enterprise AI applications.

Mythos Model Family: Anthropic’s “Invisible Arms Race”

The Mythos series plays a special role in Anthropic’s product matrix. Unlike the Claude series, Mythos is not directly aimed at end users but serves as an “internal benchmark model” for Anthropic’s internal model training—new models are first tested against Mythos before release.

The completion of training for the new Mythos version suggests that the capability ceiling of Anthropic’s next-generation Claude models has been determined. If directly named Sonnet 5 and released to the market, it indicates the company is skipping internal testing phases and directly opening new capabilities to users.

Accelerating AI Model Release Pace: From Quarterly to Weekly

Looking back at Anthropic’s release trajectory this year, the speed of model iteration has clearly accelerated:

• January 2026: Claude 4.5 Sonnet released
• March 2026: Fable 5 and Mythos 5 launched simultaneously
• May 2026: New Mythos completed training (Andrew Curran’s message)
• June 2026: Sonnet 5 expected to go online (this week)

ProCap Financial Chairman Anthony Pompliano posted on X the same day, noting that Mythos models broke into U.S. national security domain (NSA) classified systems within hours, indicating that Artificial General Intelligence (AGI) is approaching. “Current AI technology not only surpasses human capabilities but is also self-training and improving at an incomprehensibly fast rate.”

The competition in the AI model market has extended from OpenAI’s GPT series to Anthropic’s Claude family. If Sonnet 5 is released as scheduled this week, the market will focus on its reasoning ability, multi-modal support, and whether API pricing continues the previous “value-for-money” strategic route.