MarkTechPost
AI news source. Articles are auto-selected and adapted by Hamidun News editors.
Latest publications
Google Gemma 4, NVIDIA and OpenClaw: Local AI Agents With No Per-Token Bill
Google and NVIDIA are pushing Gemma 4 on RTX, Jetson and DGX Spark so always-on AI agents like OpenClaw run faster and cheaper, with no cloud token bills.
Talkie-1930: Researchers Release a 13B Model With No Knowledge of the Internet or World War II
Talkie-1930 is an open 13B model trained only on English-language text published before 1931, built to study historical reasoning, data leakage and how well AI generalizes outside its era.
MarkTechPost shows how to build a lightweight VLA agent with a latent world model and MPC
A new MarkTechPost tutorial walks through building a simplified embodied agent that works from RGB frames, learns a latent world model, and plans actions with MPC.
Arcee AI releases Trinity Large Thinking, an open reasoning model for AI agents
Arcee AI has released Trinity Large Thinking's weights under Apache 2.0, betting on long agentic workflows, multi-step reasoning, and reliable tool use.
NVIDIA shows a full model-optimization pipeline with FastNAS pruning and fine-tuning
NVIDIA has published a hands-on Model Optimizer guide: one Colab notebook covers ResNet20 training, FastNAS pruning to a FLOPs limit, and accuracy recovery through fine-tuning.
TII releases Falcon Perception, a 0.6B model for text-prompted object segmentation and detection
TII has open-sourced Falcon Perception, a compact 0.6-billion-parameter vision-language model that finds and segments objects from plain text and beats SAM 3 on complex queries.
Google DeepMind lets an LLM rewrite game-theory algorithms and beat the experts
Google DeepMind showed that AlphaEvolve can rewrite the code of algorithms for imperfect-information games and find solutions that beat hand-built expert versions on most tests.
Z.AI shows how to build production-ready agent systems on GLM-5 with tool calling
Z.AI has published a detailed GLM-5 tutorial covering SDK setup, the OpenAI-compatible API, streaming, tool calling, JSON output and a multi-turn production agent.
Netflix open-sources Void, a video object-removal model that accounts for scene physics
Netflix and INSAIT have released Void as open source: a system that removes objects from video and recalculates the falls, support and other physical consequences of their absence.
How AI Is Helping Clothing Brands Design the Fashion of the Future
Algorithms already help fashion brands build collections faster, forecast trends, cut overproduction and personalize shopping, but they also raise questions about authorship and labor.
How to Build the Netflix Void Pipeline for Removing Objects from Video with CogVideoX
A detailed walkthrough shows how to set up Netflix's Void model, download the required checkpoints, prepare inputs and run object removal from video through CogVideoX.
Gladstone Institutes unveils MaxToki, an AI model that predicts cell aging
Gladstone Institutes' MaxToki learns to read a cell's trajectory over time rather than a single snapshot, estimates aging acceleration and points to targets for intervention.
TinyFish launches unified web platform for AI agents with Search, Fetch, Browser and Agent
TinyFish has combined search, page rendering, browser sessions and autonomous web workflows into one AI agent platform with a single API key and a shared credit pool.
Google adds Skills to Chrome, turning AI prompts into one-click workflows
Google is launching Skills in Chrome: Gemini users can save frequently used prompts as repeatable workflows and run them in the browser with a single click.
Google DeepMind unveils Gemini Robotics-ER 1.6 for robot autonomy and instrument reading
Google DeepMind has updated Gemini Robotics-ER to version 1.6: the robot model better understands space and multiple video streams, and can read complex industrial instruments.
MarkTechPost breaks down the full LLM training cycle, from data to deployment
MarkTechPost has published a detailed walkthrough of how LLMs are built today: pretraining on huge corpora, SFT, RLHF, reasoning optimization and production deployment.
Google unveils Gemini 3.1 Flash TTS: a speech model with fine-grained control, dialogue and 70+ languages
Google has launched Gemini 3.1 Flash TTS in preview. It reads text in 70+ languages, handles two-voice dialogue and lets developers steer delivery with audio tags.
Mem0 and OpenAI: Building a Universal Long-Term Memory Layer for AI Agents
A new tutorial combines Mem0, OpenAI models and ChromaDB to extract facts from ordinary conversations, store them by user_id, search them by meaning and feed them back into agent replies.
SmolAgents: How to Build a Multi-Agent AI System with Code Execution and Dynamic Orchestration
A SmolAgents walkthrough shows how lightweight AI agents run code, call tools, use memory and coordinate tasks within a single system.
NetKet and JAX: Building a Transformer Model for Frustrated Spin Systems
A guide shows how to use NetKet, JAX and a Transformer architecture to build a research-grade VMC pipeline for the frustrated J1-J2 spin chain.
OpenAI unveils GPT-Rosalind, an AI model for biology, genomics and drug discovery
OpenAI has launched GPT-Rosalind, a specialized model for biology and pharma that helps speed up drug discovery, genomic data analysis and experiment planning.
Google unveils Auto-Diagnose, an AI system that finds the cause of integration test failures
Google has introduced Auto-Diagnose, a Gemini 2.5 Flash-based system that reads the logs of failed integration tests and finds the root cause with 90.14% accuracy.
OpenAI GPT-OSS: Running the Open-Weight Models in Colab with MXFP4 and Advanced Inference
A new guide shows how to run openai/gpt-oss-20b in Google Colab: install Transformers dependencies, check the GPU, enable MXFP4 quantization and build a working inference pipeline.
Anthropic releases Claude Opus 4.7 for agentic coding, vision and autonomous tasks
Anthropic has unveiled Claude Opus 4.7, an update to its flagship model with notable gains in agentic coding, high-detail image analysis and long autonomous tasks.
PrismML Bonsai: How to Run a 1-Bit Model on CUDA with GGUF, JSON and RAG
A new PrismML Bonsai tutorial shows how to run Bonsai-1.7B on a GPU via CUDA and GGUF, measure speed, set up chat, strict JSON output and a basic RAG scenario.
xAI launches standalone Grok speech-to-text and text-to-speech APIs for enterprise developers
xAI has split Grok's voice stack into separate Speech-to-Text and Text-to-Speech APIs for business, with aggressive pricing, multilingual support and a push into voice applications.
NVIDIA releases Ising, the first open family of AI models for quantum-classical systems
NVIDIA has unveiled Ising, an open family of AI models for calibrating quantum processors and correcting errors, aimed at bringing useful hybrid quantum-classical systems closer.
Why LoRA Breaks in Production and How RS-LoRA Rescues Model Fine-Tuning
LoRA handles style and format well but loses signal when fine-tuning on new facts. RS-LoRA fixes this by switching to a more stable scaling formula.
OpenKB and OpenRouter show how to build a local AI knowledge base with Llama-powered search
A new tutorial explains how to deploy a local knowledge base with OpenKB, connect an open model through OpenRouter, and set up search securely without hardcoding keys.
OpenAI and Magika Show How to Build a Pipeline for File Identification and Threat Analysis
A hands-on guide shows a pipeline in which Magika identifies a file's real type from its bytes, while OpenAI helps interpret the result and flag potential risks.