Anthropic Releases Fifth-Generation AI Models
Anthropic has announced the release of two new fifth-generation AI models: Claude Fable 5 for general use and Claude Mythos 5, which is initially available only to selected partners for specialized areas such as cybersecurity. Both models share the same base model but differ in safety guardrails and application focus.
Fable 5: General-Purpose Powerhouse
Claude Fable 5 is designed for general use and ships with conservative safety guardrails. According to Anthropic, Fable 5 outperforms all of the company's previous models, achieving top scores in benchmarks for programming, image processing, and complex data analysis. The company claims state-of-the-art results in nearly all benchmarks tested, with the gap widening on long, complex tasks.
Benchmark Performance
On SWE-Bench Pro, a benchmark for solving real software engineering tasks from public GitHub repos without assistance, Fable 5 achieved 80.3 percent. This compares favorably to Claude Opus 4.8 at 69.2 percent, GPT 5.5 at 58.6 percent, and Gemini 3.1 Pro at 54.2 percent. On Cognition's FrontierCode benchmark, which tests demanding coding tasks under production standards, Fable 5 scored 29.3 percent, while Claude Opus 4.8 managed 13.4 percent and GPT 5.5 got just 5.7 percent.
Real-World Impact
Payment processor Stripe reported that Fable 5 compressed five months of engineering work into days. In a Ruby codebase with 50 million lines, the model completed a migration in one day that would have taken a full team over two months. Anthropic also claims Fable 5 is more token-efficient than earlier Claude models, posting the top score among all frontier models on FrontierCode at medium effort.
Mythos 5: Specialized for Science and Security
Claude Mythos 5 drops the safety restrictions found in Fable 5 in areas like cybersecurity and is reserved for a small group of partners. It shows strong performance in drug design and operates largely autonomously in genomics research. The model is no longer in preview and is now available to selected partners for specialized applications.
Pricing and Availability
The new models come at a price of $10 per million input tokens, making them nearly twice as expensive as the Claude Opus 4.8 model. Token efficiency is still to be determined, but early indications suggest improvements over previous generations.
Comparison with Competitors
Anthropic provided comparisons across 15 benchmarks, showing Fable 5 and Mythos 5 outperforming Claude Mythos Preview, Claude Opus 4.8, GPT 5.5, and Gemini 3.1 Pro. Asterisked values indicate where Fable 5 scores slightly lower than Mythos 5 due to safeguard fallbacks to Opus 4.8.
Implications for AI Development
The release of Claude Fable 5 and Mythos 5 marks a significant step in AI capabilities, particularly in coding and scientific research. With real-world applications already demonstrating dramatic efficiency gains, these models could accelerate progress in software engineering, drug discovery, and genomics. However, the higher cost and restricted availability of Mythos 5 may limit immediate widespread adoption.
This article is based on reporting by The Decoder. Read the original article.
Originally published on the-decoder.com