Claude Fable 5.2 and Claude Mythos 5.2 are upgraded models aimed at demanding, open-ended technical work, with gains in coding, reasoning, vision, science, and robotics. They share the same underlying model, but Fable includes strong cybersecurity and biology safeguards and is broadly available; Mythos access is restricted to organizations doing beneficial cybersecurity or biomedical research. The models generate outputs up to 30% faster than version 5.1 and cost $8 per million input tokens and $40 per million output tokens, with an optional fast mode costing twice as much. The release recommends Opus 5.5 for routine tasks, reserving Fable or Mythos for more complex work, or using Fable to advise or coordinate smaller models.
Benchmark results put Fable 5.2 ahead of Opus 5.5 on most tested capabilities, including agentic coding, scientific research, robotics, search, and chart reading. It reached 49.1% on FrontierCode at its highest effort, while matching Opus’s best coding score at roughly half the task cost at high effort. Reported demonstrations include migrating a production service from Python to Rust, designing electronic devices, advancing physics and robotics work, and building a bootable operating system with more than 1,000 subagents over eight days. Safety tests found low assessed high-stakes misalignment risk; Fable’s cyber and biology safeguards remain robust while causing fewer unnecessary blocks, and new protections target distillation attacks.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Researchers tested 16 LLMs on all 55 TIS-100 puzzles, giving each one attempt and time to refine its code. The models found 133 new performance tradeoffs across 18 puzzles and set three outright cycle records, though human players improved on most solutions within days. (tis-100.wincon.dev)
Australia is considering a new criminal offence for AI companies that fail to take reasonable steps to prevent harm caused by their models. The government is also weighing an AI safety watchdog with power to impose significant fines, following a hack of a Medicare statistics portal involving OpenAI. (abc.net.au)
AI is increasingly becoming a gateway through which people find information. Audiences still resist having AI produce their news. (niemanlab.org)
Nvidia holds a central position in the AI industry, making disruption to the company potentially catastrophic for the wider ecosystem. The race for advanced AI is also intensifying economic and geopolitical conflict, while concentrated corporate power and infrastructure face growing political resistance. (france24.com)
LLM inference can require reading about 26 GB of model weights to generate each token from a 13-billion-parameter model. The explanation shows how model size drives memory bandwidth demands during text generation. (youtube.com)
Deletion Tuning matches full fine-tuning across three LLM post-training methods by zeroing fewer than 0.05% of model weights. Its compact masks also preserve prior-task accuracy better than full fine-tuning, LoRA, and random sparse tuning. (arxiv.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.