Archive, page 4
JEV-35B is an Apache-2.0 text-classification model built on Qwen3.5-35B-A3B, with NVFP4 and GLM5.3-Flash variants also published. A GuruSearch recommendation and search demo is live, and it has 6.3k downloads.
AWS explains running OpenAI's GPT-6 Sol, Luna and Astra models on Amazon Bedrock via global cross-Region inference from Thailand, Malaysia, Singapore, Indonesia and Taiwan, using Converse and OpenAI-compatible APIs.
Pendo released a game set in liminal spaces that teaches players how AI agent training works.
MiniCPM-V 4.7 is the latest in OpenBMB's vision-language series, including a 35B-A3B variant that is its first MoE model, plus a 1B version and GGUF quantizations. It handles single-image, multi-image and video understanding.
GitHub Copilot's local sandboxing feature is now generally available, but network access remains enabled by default.
Tabby released Tabby 2.0, adding Tabby Talk, an AI agent that answers financial questions and completes bookkeeping tasks for users.
Meta launched Muse Wonder and a second model from its revamped Superintelligence Labs, claiming performance that rivals industry leaders. Zuckerberg says Meta has reached the AI frontier.
Proba NTree is an AI agent that produces traceable fertilizer emission data, letting users track and verify agricultural emissions figures through an automated workflow.
A technical guide for deploying decision models with llama.cpp on Amazon SageMaker AI, serving OpenAI-compatible chat completions and extending runtime with SemIf to return probabilities over declared options on CPU or GPU.
A guide to deploying one large language model across two GPU nodes on Amazon EKS using Ray LLM Deep Learning Containers, with separate prefill and decode nodes behind one OpenAI-compatible endpoint.
Boole-1.7B is a 1-bit version of Qwen3-1.7B with 247 MB packed weights, 13.9x smaller than BF16 and scoring 85.9% of the BF16 model across ten tasks. It can be run locally.
Agent Role Contracts v0.5.0 is a tool for checking AI-agent workflows against defined role contracts to catch misbehavior before deployment.
A Reddit post notes Claude Code's Projects update splits one message into parallel tasks, each on its own branch, running while the laptop is closed, though users focus on usage limits.
Google's Gemini Notebook, formerly NotebookLM, released iPhone and iPad version 1.61.7 with bug fixes and improvements to its research and study tools.
Developer RafaelNGP released a native Xbox Game Pass app for jailbroken PS5s, enabling Xbox Cloud Gaming with four controllers and 4K upscaling via the PS5 GPU, built using Claude.
River AI opened an inference service priced 20% below DeepSeek and Z.ai rates, adding pressure to the already competitive AI model price war.
Alibaba's Qwen launched QwenGyre, an online reinforcement learning framework for long-term agent tasks, per KuCoin.
A new app lets you type any topic, person or moment and produces a researched two-host podcast with sources and artwork that you can interrupt.
Mistlabs, founded by a former Plaud exec, raises $26M pre-A and takes its Ambi AI ambient listening system out of beta on Apple devices.
After a pilot, The RealReal is rolling out its AI shopping agent to all 45 million members, giving them automated help finding resale luxury items.
A ComfyUI extension runs Nvidia RTX Video Super Resolution on long videos with disk-backed processing and low RAM use, supporting upscaling, denoising, deblurring, NVENC encoding and audio preservation.
DeepSeek V4.1 Flash sees setup guides, spec and price breakdowns, plus MiMo-V2.6-Pro comparisons and AMD MI355X AA-AgentPerf results.
GitLab 19.4 introduces goal-driven AI agents, hosting for open-weight models, and AI cost controls for development teams.
Decodo released a Web Data API that lets AI and RAG teams collect and structure web data for retrieval pipelines.
whatmodel.sh is an open-source tool where you enter your RAM and GPU to see which local models you can run, their rank and run command, and compare them with API models.
Indonesian telco Indosat launched an AI platform offering 101 catalogue entries for enterprise and developer use.
Crossmint released a toolkit that lets AI agents make and receive payments programmatically.
Xiryn is an open-source personal AI agent that runs locally on your computer to handle tasks on your behalf.
Decepticon is an autonomous red team agent that integrates Claude and Codex to automate offensive security testing.
Version 2.14.0 of Mistral Vibe (Le Chat) for iPhone, an AI agent for long-horizon tasks, brings various fixes and improvements.
Microsoft Copilot can now create and edit Word, Excel and PowerPoint files, while Copilot themes in Outlook are retired starting mid-October 2026 after user pushback.
Mistral unveils ML4, an open-weight model that it says gives it an edge over Chinese rivals, as it competes with US labs.
xAI releases Grok 4.7, with pricing and benchmark comparisons against Anthropic's Claude 5.5 published alongside it.
OpenAI raises the default context length of GPT-6.1 Sol to 500,000 tokens.
Atticus is a Mac voice assistant: hold Control-Option-Space, speak or type a task, and it runs it, sending simple tasks to cheaper handling. Built by a 9th-grade developer and shown as Show HN.
Decision 3.0 is a family of five open multimodal models from 0.8B to 27B parameters that answer choice, yes/no and score questions about text and images in one call without generating text. Released under Apache-2.0 on Hugging Face.
Mira Tours is an AI walking tour guide app launched on Product Hunt, where it is open for discussion and linked for access.
Claude Haiku 5.5, Anthropic's smaller model, reaches GitHub Copilot, with pricing and best-use guidance published. The source text does not give full pricing or context details.
Takes for Mac is a Mac app for recording talking-head videos, where Claude Code handles the editing. Launched on Product Hunt with discussion and a link.
SupoClip is an open-source AI clipper that cuts long videos into short viral clips. Launched on Product Hunt with a link and discussion.