Anthropic Launches Claude Fable 5.1 and Mythos 5.1 Amid IPO Speculation

On September 1, 2026, Anthropic unveiled its latest AI models, Claude Fable 5.1 and Claude Mythos 5.1, designed to enhance performance in coding and knowledge-intensive tasks. This launch comes as the company gears up for a potential initial public offering (IPO), with market predictions suggesting a 90% chance of completion by the end of October.

New Features and Pricing

Claude Fable 5.1 is now available broadly through the Claude API, while Claude Mythos 5.1 is restricted to select users within the Project Glasswing program. Both models support a context of one million tokens and offer significant improvements over their predecessors, including mid-conversation effort adjustment and content provenance tracking for verifying output origins.

The pricing structure remains unchanged, with costs set at $10 per million input tokens and $50 per million output tokens. However, cache read costs have dropped significantly, making typical workloads 25% cheaper and complex tasks up to 45% cheaper.

Strategic Implications

Anthropic's move to restrict new API accounts from editing conversation history aims to prevent competitors from distilling their models, addressing concerns over intellectual property theft. This follows previous reports of significant distillation attempts by various entities, including accusations against specific labs for copying Anthropic’s technology.

As the company prepares for its IPO, which could value it at approximately $965 billion, the launch of these advanced models signals Anthropic's commitment to maintaining a competitive edge in the rapidly evolving AI landscape. The upcoming public offering will require the company to disclose detailed financial information, potentially impacting its valuation stability.

New Developments in Claude Fable 5.1

• Claude Fable 5.1 has achieved the highest score of 66 on the Artificial Analysis Intelligence Index, surpassing over 192 competing models.
• The model was released on September 1, 2026, and costs approximately $3.69 per task, which is about 20% more than its predecessor, Fable 5.
• Despite the increase in per-task costs, cache read prices have been reduced by 75%, from $1 to $0.25 per million tokens.
• Fable 5.1 can generate up to 140 million output tokens for complex tasks, significantly increasing costs for demanding workloads.
• The model supports a 1 million-token context window, enhancing its capabilities in long-horizon agentic coding and multi-step research.
• Fable 5.1 outperforms Claude Opus 5.1, which was launched in July 2026, indicating Anthropic's rapid development pace with three major model launches in three months.
• The model is available in Pro, Max, Team, and Enterprise tiers, targeting the premium segment of the market.