AI News
Stay updated with the latest developments in artificial intelligence
Product Launch Aug 12, 2026 1 min readGemini 3.1 Pro Listed in Preview With Length-Tiered Pricing
Google lists Gemini 3.1 Pro as a preview model, priced $2/$12 per 1M tokens for prompts up to 200k and $4/$18 beyond that.
Open Source Aug 11, 2026 1 min readNVIDIA Puts a 30B Agentic Model on a Single Laptop GPU
Nemotron 3.5 Lightning is a 30B mixture-of-experts model with 3B active parameters that runs on one GPU, shipped alongside NeMo Switchyard, an open-source router that picks a model per agent step.
Security Aug 10, 2026 1 min readOpenAI Ships GPT-5.6-Cyber Behind Identity Checks and Hardware Keys
A GPT-5.6 Sol branch trained to refuse less on dual-use security work, gated behind the Daybreak Red programme, priced at $12.50 / $75 per million tokens.
Product Launch Aug 5, 2026 1 min readMeta Ships Muse Spark 1.2 and Its First Coding Agent
Meta released Muse Spark 1.2 on August 5, 2026 alongside Muse Code, its first terminal coding agent. The model landed at rank 4 on LMArena's text leaderboard - the highest debut for any Meta model tracked on this site.
Product Launch Aug 3, 2026 1 min readAlibaba's Qwen3.8-Max Reaches 2.4 Trillion Parameters
Alibaba launched its largest model on 3 August 2026: a 2.4T-parameter mixture-of-experts system with 95B active per query, natively accepting text, images and video.
Open Source Jul 31, 2026 1 min readKorea's Largest Model Goes Apache 2.0
LG AI Research published K-EXAONE 2.0 - 750 billion parameters, the largest foundation model built in Korea - to Hugging Face under Apache 2.0, permitting unrestricted commercial use.
Product Launch Jul 24, 2026 1 min readMidjourney V8.2 Becomes the Default - and There Is No V9
V8.2 took over as Midjourney's default model on 24 July 2026, focused on aesthetics and personalisation. Despite recurring speculation, V9 has not been announced.
Product Launch Jul 23, 2026 1 min readFLUX 3 Turns an Image Model Into an Image, Video, Audio and Robot Action Model
Black Forest Labs unveiled FLUX 3 on 23 July 2026: one architecture jointly trained on images, video and audio, extended to predict robot actions - with most of it still behind early access.
Open Source May 20, 2026 1 min readCommand A+ Runs a 218B Model on Two H100s Under Apache 2.0
Cohere's first fully Apache-licensed frontier model is a 218B sparse mixture-of-experts with 25B active parameters, deployable on as few as two H100 GPUs.
Product Launch Mar 26, 2026 1 min readSuno v5.5 Adds Voice Cloning and Personal Custom Models
Suno v5.5 shipped three personalisation features: Voices for capturing your own singing voice, Custom models trained on your own catalogue, and My Taste.
Open Source Apr 5, 2025 1 min readLlama 4 Scout: 10M Token Context in an Open-Weight Model
Meta's Llama 4 Scout pairs a 10-million-token context window with a mixture-of-experts design of 17B active and 109B total parameters.
Showing 145–155 of 155 stories