Model Releases

Tencent Hunyuan has introduced Hy4 preview, an open model with 770B total parameters, 4.9B activated parameters, and a 1M-token context window. The release also highlights integrations with game engines and MCP-oriented tool workflows.

Alibaba’s DAMO Academy and clinical partners have developed DAMO LiON, an AI system that assists doctors in detecting small malignant liver lesions on contrast-enhanced CT scans. In a two-month prospective real-world trial, it helped identify 15 previously overlooked malignancies.

Ox Alpha has drawn intense attention since its free release on OpenRouter, yet the team behind it remains anonymous. Speculation has ranged from Z.ai’s GLM models to an unreleased Microsoft MAI system, with no confirmed answer.

Motif 3 is a decoder-only sparse MoE language model with 314B total parameters and 13.2B activated per token. Its design combines GDLA, training-stability tricks, and multi-teacher post-training to improve reasoning, coding, and long-context performance.

DeepSeek-V4-Flash is now available through an official public beta API. Its architecture and scale remain unchanged, while post-training appears to be the main driver behind the reported gains in agent and coding benchmarks.

Kimi K3 is presented as a 2.8T-parameter Mixture-of-Experts model with 104B activated parameters per token, native vision support and a one-million-token context window. The paper positions it as a frontier-level open-weight model for long-horizon coding, agentic execution, reasoning, knowledge and vision tasks.

Ant Bailing has released Ling-3.0-Flash, a new native hybrid reasoning model with 124B total parameters and 5.1B active parameters per computation. The model focuses on agent workflows, long-context processing, and a better balance between capability and inference cost.

Anthropic’s Claude Opus 5 is positioned as a lower-cost rival to Fable 5, with strong benchmark results and impressive developer demos. The more important shift may be its ability to verify its own work and rely less on rigid system prompts.

Anthropic has introduced Opus 5, a new heavyweight model that is smaller than Fable 5 but cheaper, less restricted, and stronger on several announced benchmarks. The release also brings a beta fallback feature designed to reduce hard stops from safety classifiers.

Anthropic has introduced Claude Opus 5 as a daily-use model that aims to approach Claude Fable 5’s frontier intelligence at roughly half the cost. The announcement highlights gains in coding, automation, knowledge work and scientific workflows, while noting that cybersecurity remains a weaker area versus Mythos 5.

Moonshot AI has introduced Kimi K3 with 2.8 trillion parameters, a MoE architecture, a 1-million-token context window, and native multimodal capabilities. If its full weights are released as planned, it could become the first open-source model in the 3-trillion-class range.

OvisOCR2 converts document page images directly into Markdown in natural reading order, covering text, formulas, tables, and visual regions. Its reported results suggest that compact end-to-end parsers are becoming serious competitors to traditional pipeline-based OCR systems.

TechCrunch, citing the Financial Times, reports that Moonshot AI’s upcoming Kimi K3 could become China’s largest open-weight AI model and may narrow the gap with Anthropic’s Opus 4.8.

A new arXiv paper introduces VAIOM, a decoder-only Transformer designed for probabilistic next-return modeling on one-hour foreign-exchange bars. The model keeps financial observations continuous at the input while using discrete return buckets for training and evaluation.