Cactus releases Needle, a 26M parameter function calling model
Cactus Compute released Needle, a 26 million parameter model specialized for function calling, claiming it outperforms much larger models. The model is open-source with a demo on Hugging Face and code on GitHub.
Entities
Related
Tencent Hy Team releases Hy3, a 295B-parameter MoE model
Tencent's Hy Team released Hy3, a 295-billion-parameter Mixture-of-Experts (MoE) language model with 21B active parameters and a 3.8B MTP layer. The model outperforms similar-sized models and rivals open-source models with 2-5x parameters. It is available on Hugging Face in full (598GB) and FP8 quantized (300GB) versions.
Colibrì 744B-Parameter Model Runs on a Laptop
A new 744-billion-parameter model named Colibrì has been released, capable of running on a laptop. The model's name and parameter count suggest a focus on efficiency despite its large size.
Zer0Fit: MCP server for zero-shot ML with Google TabFM & TimesFM
Zer0Fit is an MCP server that wraps Google's TabFM and TimesFM foundation models, enabling zero-shot classification, regression, and forecasting via a local LLM. It allows users to perform ML tasks without training or tuning, connecting to Open WebUI, Claude Code, or Codex. Built for developers and researchers who want quick, local ML inference.
OpenAI releases GPT-5.6 with Sol model, claims to outperform Claude Fable
OpenAI released GPT-5.6 on July 10, 2026, featuring a new Sol model that reportedly surpasses Anthropic's Claude Fable on benchmarks. The release was covered by Fireship on YouTube, noting the timing and performance claims.
Hugging Face integrates native-speed vLLM backend into Transformers library
Hugging Face has upgraded the vLLM pip package to support native-speed inference directly within the Transformers library. The update enables seamless use of vLLM as a backend for Transformer models, leveraging its optimized serving capabilities. This integration allows developers to run models at native speed without switching frameworks.