TechTurco WhatsApp ChannelOne-way channel: your number stays private
Artificial Intelligence
Claude Fable 5.2 and Mythos 5.2 announced: Developed an operating system from scratch
Claude, one of the prominent artificial intelligence (AI) platforms, has been updated with new models. Anthropic announced Fable 5.2 and Mythos 5.2, which run 30 percent faster than their predecessors.
Contents
Two months after the release of version 5.1, Claude's creator Anthropic officially introduced its flagship artificial intelligence (AI) models, Claude Fable 5.2 and Mythos 5.2. According to information shared by the US-based company, the new models include notable improvements in coding, reasoning, and scientific research. Capable of generating responses up to 30 percent faster than the previous generation, the models also operate at lower operational costs.
Major leap in benchmark tests: Coding, robotics, and science
According to benchmark results published by Anthropic, Fable 5.2 scored 49.1 percent on the FrontierCode 2.0 test. This marks a direct increase compared to Fable 5.1's score of 32.4 percent. In scientific research tasks, the model's performance rose from 52.6 percent to 74.6 percent.
The largest percentage increase was recorded in robotics simulations. The robotics simulation score, which stood at 27.6 percent in Fable 5.1, reached 64.9 percent with Fable 5.2.
Reklam
| Benchmark / Task Area | Claude Fable 5.1 | Claude Fable 5.2 |
|---|---|---|
| FrontierCode 2.0 (Coding) | 32.4% | 49.1% |
| Scientific Research Tasks | 52.6% | 74.6% |
| Robotics Simulations | 27.6% | 64.9% |
Built an operating system by managing over 1,000 agents
Tests conducted by Anthropic engineers evaluated the model's capacity for large-scale tasks. Tasked by an engineer with developing an operating system from scratch, Fable 5.2 coordinated more than 1,000 artificial intelligence (AI) agents over an eight-day period. The process resulted in a functional operating system capable of booting on a laptop.
The company also stated that Fable 5.2 solved two previously unresolved open problems regarding black holes and contributed to formalizing the mathematical proof of Fermat's Last Theorem within 30 hours.
Mythos 5.2 leads in detecting security vulnerabilities
Fable 5.2 and Mythos 5.2 share the same underlying architecture, but the Mythos model features more flexible safety filters designed for sensitive cybersecurity and biomedical research. Working on identifying security vulnerabilities, the model achieved an 89.6 percent success rate on the CVE Rediscovery benchmark, compared to 78.3 percent for Mythos 5.1.
Reklam
Be the first to know
Join our channel and get key developments straight on your phone.
TechTurco Telegram ChannelMute or leave any time
Prices dropped, access policies diverged
API usage pricing was also reduced alongside the announcement. The updated rate schedule for both models is as follows:
- Input token price: $8 per 1 million tokens (Previously $10)
- Output token price: $40 per 1 million tokens (Previously $50)
- Fast mode: Generates responses up to 2.5 times faster but is billed at double the standard rate
A distinction remains regarding access to the models. While Claude Fable 5.2 remains accessible to general users and independent developers, Mythos 5.2 continues to be under strict oversight. Additionally, using Mythos 5.2 requires approval from Anthropic's cybersecurity or life sciences verification programs. Enterprise customers can access this model's capabilities through Claude Security.
Reklam
This article was translated from Turkish. Read the original article in Turkish.
More in Artificial Intelligence
