Models
New LLMs, multimodal and open-weight model releases and benchmarks
- Block introduces speech model Dessa to recognize emotions and sarcasm Models
Block, the company behind Square and Cash App, has introduced Dessa, a speech model that processes audio directly instead of converting it… - Coding performance close to Opus 5.5, 3-trillion-class LLM that also makes 3D "Kimi K3.1 Pro". Open-weighting by November 4 Models
An improved version has already appeared for "Kimi K3," which was the first to reach the 3T (3 trillion) parameter class as an open large… - "Claude Mythos 5" targets real developers regardless of signs of harm — Anthropic and AISI release analysis results Models
Anthropic and the UK AI Safety Institute (AISI) released an analysis report regarding an incident where "Claude Mythos 5" targeted real… - AI model "Griffin" appears, which 48% of participants in video calls misidentified as human, reproducing face and body movements in video from a single image Models
AI company Tavus announced the AI model "Griffin," which interacts with humans using real-time video and audio. In an experiment using the… - Mistral releases new model 'Le Chonk' Models
Mistral has released its new AI model 'Le Chonk', a trillion-parameter model. It aims to be the best open-weight offering outside of China. - Reflection introduces US-originated open-weight model "Beam" Models
Reflection announced its first open-weight model "Beam" on the 5th (US time). It is a Mixture-of-Experts (MoE) model with 501B total… - Reka Releases Rho-1: A 19B Omni-Reasoning Model That Understands, Generates Video and Outputs Robot Actions in One Models
Reka has released Rho-1, a 19B omni-reasoning model that processes and generates text, images, video, and robot control actions in a…