The AI landscape just got a seismic shake-up. Anthropic has officially unveiled two new flagship models—Claude Fable 5 and Claude Mythos 5—pushing the boundaries of what large language models can achieve, while introducing nuanced access rules and premium pricing to match their capabilities.
Dual Model Launch: Fable 5 for All, Mythos 5 for Trusted Partners
Claude Fable 5 is now fully accessible to all users across every Anthropic channel, marking the company’s most powerful public release to date. In contrast, Claude Mythos 5 is initially limited to Glasswing partners, with plans to expand access through Anthropic’s Trusted Access program in the future. Both models share the same underlying architecture, but Mythos 5 removes certain safety restrictions to unlock more specialized, high-risk use cases.
Breakthrough Performance: Redefining AI Capabilities Across Domains
Claude Fable 5 outperforms all previous publicly released Anthropic models, leading the pack in multi-domain AI benchmark tests. Its real-world applications are nothing short of transformative:
- Software Engineering: Cut the time to migrate a 50-million-line Ruby codebase from two months of manual work to just one day.
- Knowledge Work: Claimed the highest score on the Hebbia Finance Benchmark, with sharpened document reasoning and chart interpretation skills.
- Visual Tasks: Set a new state-of-the-art (SOTA) bar, extracting precise values from scientific charts, reconstructing web source code from screenshots, and even beating Pokémon Red with minimal visual prompts.
- Memory & Context: Maintains focus while processing millions of tokens, and delivers performance 3x better than Claude Opus 4.8 in the strategy game Slay the Spire.
- Scientific Applications: Demonstrates strong proficiency in drug design and other complex scientific tasks.
For restricted-access Claude Mythos 5, the gains are even more specialized. Built on the same base as Fable 5 but with certain safety restrictions removed, it can accelerate key drug design workflows by approximately 10x. In blind tests, scientists preferred Mythos 5’s novel scientific hypotheses 80% of the time; some of these hypotheses are already moving into experimental validation, including one about a new protein mechanism in E. coli that has been confirmed by an independent laboratory.
Pricing: Premium for Premium Performance
Both models come with a premium price tag: $10 per million input tokens and $50 per million output tokens. This is double the cost of Claude Opus 4.8, and compares to GPT-5.5 at twice the input price and two-thirds higher output price. Notably, this pricing is less than half the cost of the Claude Mythos Preview.
Temporary Free Access for Subscribers
To let users experience Fable 5’s capabilities firsthand, Anthropic is offering free access to the model through June 22. This benefit applies to Pro, Max, Team, and seat-based enterprise subscribers. Starting June 23, Fable 5 will be removed from subscription plans, and users will need to consume their usage credits to access it. Anthropic has stated that once capacity allows, it plans to reintegrate Fable 5 into standard subscription benefits.
Safety Guardrails: Balancing Power and Responsibility
Given the model’s advanced capabilities and associated safety risks, approximately 5% of conversations will trigger Fable 5 to downgrade its response to Claude Opus 4.8. Additionally, Anthropic has imposed a strict ban on using Fable 5 to develop new large language models.
Industry Context: AI Race and Calls for Caution
The launch comes on the heels of an Anthropic report highlighting the accelerating pace of AI self-iteration, which prompted the company to call for a global pause on AI technology development. Meanwhile, industry rumors suggest OpenAI is set to release GPT-5.6 later this week. AI expert Andrej Karpathy praised Fable 5 as an exciting, leap-forward update, noting its improvement is comparable to the jump from Claude 4 to 4.5.
As Anthropic pushes AI capabilities forward with Fable 5 and Mythos 5, the launch underscores the dual realities of the current AI race: unprecedented innovation paired with growing calls for responsible development. For users, the choice between premium performance and accessible pricing, along with the promise of future expanded access, offers a glimpse into the evolving landscape of AI tools. For the industry, these models serve as both a benchmark of progress and a reminder of the need to balance power with caution.