Anthropic released new AI models Fable 5.1 and Mythos 5.1, aiming to reduce costs and false positives. Fable 5.1 is widely available via cloud platforms and the Anthropic API, while Mythos 5.1 is rest…
Anthropic released Claude Fable 5.1 and its restricted twin Mythos 5.1, keeping the headline token price at $10 per million input and $50 per million output while cutting cache reads by 75% to $0.25 p…
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, with Fable 5.1 scoring 52.6% on Terminal-Bench-Science 0.1, more than doubling Fable 5's 24.7%, and 55.8% on Terminal-Bench 4.…
A developer's independent analysis of reasoning prefills on open models finds that Kimi K3 shows a significant accuracy increase on Humanity's Last Exam questions when prefilled with Opus 4.8 reasonin…
OpenAI has opened a limited preview of OpenAI Ultrafast, an API tier serving GPT-5.6 Sol at up to 750 output tokens per second on Cerebras wafer-scale silicon, roughly 14 times faster than its Standar…
Cerebras and OpenAI launched Ultrafast Mode, a new service tier in the OpenAI API powered by Cerebras that delivers GPT-5.6 Sol at up to 750 output tokens per second with no quality compromise. In Cer…
Three startups building AI mathematicians—Harmonic, Axiom Math, and Math Inc—raised over $580 million in the last twelve months, with Harmonic at a $1.45 billion valuation and Axiom Math jumping from …
Thinking Machines Lab released Inkling-Small, an open-weight Mixture-of-Experts transformer with 276 billion total parameters and 12 billion active, designed to deliver performance comparable to its l…
OpenAI is optimizing its GPT-5.6 model for the GPQA Diamond benchmark, while Anthropic is optimizing Opus 5 for the Humanity's Last Exam benchmark, with GPT-5.6 winning on GPQA Diamond and Opus 5 winn…
Researchers introduced PoTRE (Poly-Topological Reasoning Ensembles), a heterogeneous framework that decouples inference into four agents to improve complex reasoning in large language models. PoTRE ac…
Poetiq claims its Recursive Self-Improvement (RSI) Metasystem has autonomously set state-of-the-art results on six diverse benchmarks, including outperforming Muse Spark 1.1 within 48 hours of publica…
A new benchmark called Humanity's Last Exam, designed to stump frontier AI models, is now available as a 10-question randomized quiz called The Adult SAT. The quiz lets users compare their scores agai…
The Center for AI Safety, with world experts, created Humanity's Last Exam (HLE), a benchmark of over 2,500 expert-level questions across disciplines to test AI reasoning. Even top models like GPT, Ge…
Two AI judge models scored an agent's answer 0.85 based solely on final answer matching, but the agent never retrieved the required Confluence page, leading to a trace-based score of 0.000. This expos…
Google has made Gemini 3.5 Flash generally available, offering faster performance and improved agentic and coding benchmarks over its predecessor, but developers migrating from the preview version mus…
Andon Labs co-founders Lukas Petersson and Axel Backlund are stress-testing frontier AI models by giving them real-world control over inventory, wallets, and physical storefronts, revealing unexpected…