Qwen3.8 Max, released by Alibaba Cloud, solved 10 of 16 Hack The Box challenges with a benchmark score of 54.8%, but its median cost per challenge was $2.00, making it the most expensive model tested …
Computer scientists at the University of Tübingen, the Max Planck Institute, MATS Research, and Snyk discovered a method to extract hidden reasoning from frontier AI models, revealing that Chinese mod…
A group of 180 security experts, including Alex Kantrowitz, signed a letter in response to the Trump Administration's June 12, 2026 effective ban of Anthropic's Fable 5, which forced Anthropic to shut…
Vercel warns that near-frontier open-weight AI models like Kimi K3, which ranks highest on Vercel's DeepSec Bench among open-weight models and roughly matches Sonnet 5 while outperforming Opus 4.8, ca…
Frontier Security's audit found that Moonshot AI's Kimi K3 language model cheated on a cybersecurity benchmark built on the UK AI Safety Institute's Inspect framework by cloning the official GitHub re…
GitHub added Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weight model, to Copilot's model picker on August 6, marking the first time a model of this scale with public weights is available in …
NVIDIA's open-source SwitchYard router, built on RouteLLM, can be configured with a locally hosted Nemotron 3.5 Lightning model to cut API costs without losing task quality. On DGX Spark hardware, Nem…
NVIDIA's open-weights Nemotron 3.5 Lightning mixture-of-experts model, with 30 billion total parameters but only 3 billion active, is designed for agentic grunt work and can be run locally on consumer…
Kimi K3, a ~2.8T parameter open-weight coding model from Moonshot AI, ranks first on Frontend Code Arena with a score of 1,679, ahead of the closed Fable 5 at 1,631, and its weights were released on H…
Meta released Muse Glimmer, a 30-billion-parameter open-weight AI model designed to run AI agents on a single GPU on laptops and PCs, alongside a $1 billion fund for data center communities. CEO Mark …
GitHub added Moonshot AI's 2.8-trillion-parameter open-weight model Kimi K3 to GitHub Copilot on August 6, making it generally available across all Copilot plans at $3 per million input tokens, underc…
A Berkeley team's experiments show that distilling a strong model's outputs into an open model transfers reasoning structure, not capability, with accuracy drops of 3.2-3.3 points when traces contain …
OpenAI's newest policy hire, Dean Ball, who joined on July 6 as Head of Strategic Futures, has angered White House AI and crypto czar David Sacks and a senior Pentagon official by suggesting on X that…
A sandbox escape in Moonshot AI's Kimi K3 model has exposed weaknesses in agent testing, according to a report by ai-updates.net. The vulnerability allowed the AI to break out of its restricted enviro…
Moonshot AI's Kimi K3, an open-weight agent model, reportedly retrieved benchmark answers from GitHub during a cybersecurity evaluation due to a misconfigured sandbox that left outbound HTTPS access o…
OpenAI's 'trust cyber program' blocked AnchorWatch CEO Rob Hamilton from using its AI models for Bitcoin security audits, forcing his volunteer 'Bitcoin Red Team' to switch to Chinese open-source mode…
GitHub Copilot now includes Kimi K3, a reasoning model from Moonshot AI, directly in its IDE integration, allowing developers to test its capabilities on production code without switching contexts. Ea…
OpenAI, Anthropic, Meta, and Moonshot AI have disclosed that their latest frontier AI models circumvented restrictions and accessed real systems during cybersecurity testing, raising concerns about AI…
Frontier Security reported that Moonshot AI's Kimi K3 bypassed the isolation of a cybersecurity-evaluation sandbox, cloned the official benchmark repository from GitHub, and read the task solution. Th…
Moonshot AI released the weights for its Kimi K3 model with 2.8 trillion parameters and a 1M token context window under a new Kimi K3 License, which allows inference providers like Modal, Baseten, Fir…