model-release

Perceptron, a startup founded by ex-Meta scientists, has launched Isaac 0.5, an open-weight visual AI model designed to help robots perceive, reason, and act in industrial environments like warehouses and factories.

Ox Alpha is identified as the GLM 5.3 Flash model released by zAI.

Anthropic is preparing to release two new checkpoints for its Claude AI model as early as this week, signaling potential updates to its language model capabilities.

Alibaba will open-source the Qwen3.8-Flash-Next model on ModelScope at 23:00 tonight, featuring Qwen4's GDN hybrid layers and sparse attention architecture, but lacking benchmark data.

GWEN 3.8 Max is an AI model update, likely featuring improved performance or expanded capabilities.

The article speculates on the upcoming release of GLM 5.3 weights and suggests that OxAlpha may be a new GLM model variant.

Ox Alpha is a new AI model that rapidly gained adoption, driving 8% of inference on Cline within 48 hours, and is now available for free.

The article highlights the unprecedented rate of AI progress, with models becoming cheaper and more capable, emphasizing opportunities for applied AI companies and mentioning the free release of the Ox Alpha model by OpenCode.

SenseNova U1.5-Lite is released, featuring expert training and OPD distillation to create a unified model for inference, with improvements in instruction following and image generation capabilities.

Liquid AI releases DSpark draft model checkpoints for the LFM2.5 family, enabling up to 3.2x faster inference on GPUs and devices with minimal quality trade-off, and with day-one support for open-source tools like llama.cpp and SGLang.

Ling-3.0's model release provides six base checkpoints for tiny and flash variants across pre-trained, mid-trained, and WSM-merged stages, allowing users to continue training or compare model evolution.

Unsloth AI released 1-bit quantized versions of the Qwen3.8-27B AI model that run on 8GB RAM, retaining approximately 77% accuracy compared to BF16 precision.

LiquidAI has released updated 4-bit Q4_0 checkpoints for LFM2.5 models using Quantization-Aware Distillation, recovering 97% of accuracy lost to quantization while maintaining low memory and high speed.

Wispr, an AI dictation startup, raised $280 million in Series B funding at a $2 billion valuation and launched a new model called Canto to improve speech understanding accuracy.

Z.ai's GLM-5.3 model demonstrates exceptional benchmark performance, surpassing larger models like Kimi K3 and matching frontier models, highlighting advancements in post-training efficiency by Chinese AI labs.

A new version of the Huihui-ai Qwen model is available on Hugging Face, and the user is pulling it for use in analysis and services, noting previous versions were effective.

dots3-note Preview is an AI model targeting autonomous agency and dynamic real-world environments with a 512K context window.

LiquidAI released LFM2.5-VL-3B, a 3.1B local vision-language model that outperforms Gemma-4 E4B and demonstrates a significant jump in screen understanding, making on-device AI more practical.

Community discussion on the release of Qwen3.8-27B, focusing on performance comparisons, memory usage, and creative writing capabilities with previous versions like Qwen3.6-27B and Qwen3.5-27B.

Qwen 3.8 27B has been released, and the author invites users to share their experience, including which frontier model it resembles and which quantization they used.