Mistral Large 4 Preview, nicknamed "Le Chonk", scored 38 on the Artificial Analysis Intelligence Index v4.3.2, making it the highest-scoring open-weights model from outside China on the chart, tied wi…
A statement signed by 1,386 employees of frontier AI companies, including OpenAI Chief Scientist Jakub Pachocki, Anthropic CEO Dario Amodei, and Google DeepMind co-founder Shane Legg, requests that th…
A new evaluation called Duplex Cue finds that full-duplex voice agents adapt to listener contributions mid-turn far less often than humans, according to a single-model case study of 300 human-confirme…
Three researchers — Beren Millidge, CTO of Zyphra; John Schulman, chief scientist at Thinking Machines and formerly of OpenAI; and Charlie O'Neill, head of model training at Baseten — argue that gener…
AI researchers John Schulman, chief scientist at Thinking Machines, Beren Millidge, CTO of Zyphra, and Charlie O'Neill, head of model training at Baseten, debated how close the field is to recursive s…
Tinker's LoRA fine-tuning matches full fine-tuning performance on small datasets and reinforcement learning, but underperforms on larger supervised learning datasets, according to a technical primer f…
Barret Zoph, a co-founder of Thinking Machines who left the startup in January to rejoin OpenAI, has taken a position as vice president of research at Google, where he previously worked. A Google spok…
OpenUI's generative UI benchmark of 30 models found xAI's Grok 4.6 and OpenAI's GPT-5.6 Sol tied for the highest validity at 99.5%, with Grok 4.6 the cheapest frontier model at $0.0185 per task. Acros…
Harvey, the $11 billion legal AI unicorn, released Tenet, a specialized legal model built on Kimi K3, achieving nearly double the success rate on its proprietary LAB benchmark compared to the raw base…
A senior industry observer likens the 2026 AI economy to a three-body problem, where frontier labs led by OpenAI and Anthropic, Chinese open-weight models from Zhipu and Moonshot, and application comp…
A new analysis from FutureSearch forecasts that six 'neolabs'—Safe Superintelligence, Thinking Machines, Reflection AI, Ineffable Intelligence, Discovery Loop, and AMI Labs—will not achieve superintel…
Thinking Machines, founded by former OpenAI CTO Mira Murati, released Inkling on July 15, 2026, a 975-billion-parameter AI model with 66 layers and 256 experts per layer, of which only six activate pe…
A Swiss provider operating a single 64-GPU cluster in Zurich could generate about CHF 7.4 million in annual revenue and CHF 3 million in EBITDA by selling dedicated Kimi K3-class inference to banks, p…
Republican Senator Jim Banks urged the Trump administration to develop incentives for U.S. tech companies to build open-weight AI models, warning that Chinese open-weight models could be weaponized ag…
Thinking Machines' new open-weights LLM Inkling, a nearly 1 trillion parameter model, successfully rewrote several Rust functions for a developer, who applied 2 or 3 of the optimized versions to their…
Thinking Machines' 150M-parameter BDH-CQ reasoning model, which combines in-context learning with recurrent latent reasoning, achieves 29.5% pass@2 on the public ARC-AGI-1 evaluation set at a computed…
Meta unveiled Muse Glimmer, a 30-billion parameter open weight agentic model that runs on a single consumer GPU, on Monday. The model outperforms Google's Gemma4-31B and Qwen3.6-27B on benchmarks for …
DeepSeek's V4-Flash-0731 release on July 31 is 12 points more censored than its preview on China-sensitive topics, with the mean matched gap rising from +32.0 to +44.0 and the median from +33.3 to +56…
Waterr released a research preview of Monologue, a reasoning harness that gives real-time voice agents an inner monologue to prevent forgetting in long conversations. The architecture, called ORI (Omn…
Thinking Machines released Inkling-Small, a lightweight AI model that is a quarter the size of its predecessor Inkling but delivers equivalent performance, cutting GPU memory use by 75% to enable loca…