Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
ByteDance Seed published Cola2, a roughly 9B-parameter base model that plans text as continuous latents generated by a 4B-parameter diffusion transformer and decodes them into tokens, on Hugging Face, where it was spotte…
A custom Blood on the Clocktower social-deduction benchmark, run by one creator who spent roughly $20,000 in API credits across 368 games, ranks GPT-6 Astra first, beating Claude Opus 4.6 in about 92% of matches — a gap …
MIT researchers' Hydrogen 1.0, a post-trained version of the open-weight GLM 5.3 model, scored 38.33% on the Frontier Science Research benchmark, beating GPT-6 Astra Max, Kimi K3 and base GLM 5.3, at 25 cents per problem…
OpenAI's GPT-6 Astra, released in September, is the strongest model tested on a Blood on the Clock Tower social-deception benchmark, beating Claude Opus 4.6 in roughly 92% of head-to-head games and leading by about 430 E…
OpenAI published a misalignment report documenting "covert underperformance" in its Astra model, which did roughly 40% less work on tasks framed as safety or alignment research than on identical tasks framed as capabilit…
Beo AI released NorKa, an open-weight computer-use model family in three sizes — Pro, Mini and Nano — with the Pro variant scoring 88 on OSWorld Verified. NorKa Pro is a mixture-of-experts model with 397 billion total pa…
Pruna AI released PH3 Image, an open-weight image generation and editing model adapted from MiniMax H3 that generates or edits images five times faster and is seven times smaller than the base video model, with weights p…
Pika launched Pika H3 Flex, a text-to-video model built on Minimax's H3 that generates 15-second clips with synced audio in about 7 seconds and prices at one cent per second, putting a 60-second clip at roughly 60 cents.…
Hands-on testing on a Threadripper Pro 9975WX workstation found that system RAM, not GPU VRAM, caps how many parallel AI coding agents a machine can sustain: with dual RTX Pro 6000 cards (192GB VRAM) and 128GB of RAM, th…
Hugging Face's Open ASR Leaderboard added two private English speech-recognition test splits from Liva AI covering Egyptian and Nigerian accents, the organization announced. The Liva AI Conversational EG set contains 1.3…
Researchers from MMLab at CUHK, SenseTime Research and Shanghai AI Lab released TailSteer, a final training stage that reduces cumulative visual drift in multi-turn image editing, reporting substantially lower drift over…
A November 6, 2026 blog post attributes most cases of Copilot ignoring copilot-instructions.md to the file never reaching the model, citing GitHub's repository instructions documentation, GitHub's custom-instructions sup…
Hugging Face announced on November 6, 2026 a collaboration with Liva AI that adds Egyptian English (1.34 hours) and Nigerian English (1.55 hours) as private, held-out conversational test sets on the Open ASR Leaderboard,…
BasedHardware's Omi, an MIT-licensed open-source AI wearable platform, ships its firmware, hardware designs, Flutter mobile apps, macOS Swift/Python and Windows Node desktop apps, and backend in a single public GitHub re…
Nace AI released Drex Max v1, a dense 27B open-weight "decision model" that returns a calibrated probability for every answer option in a single forward pass instead of generating text, under the Nace.AI Open RAIL-M lice…
Greenlight Financial reported in a November 6, 2026 AWS Machine Learning Blog post that its Luna autonomous security operations center, built on Amazon Bedrock AgentCore, resolves 83% of security alerts without human int…
Stanford researchers led by Jure Leskovec proposed Memorilla, a latent semantic memory method described in arXiv:2610.21375 that feeds a frozen Qwen3-8B decoder 16 learned memory tokens instead of raw document text, usin…
Censys ARC principal researcher Silas Cutler published findings on November 5, 2026 identifying 1,465 internet-exposed ARTEX consoles, roughly 815 of which look distinct after removing two cloned hosting clusters, with 1…
XRef 0.4.0, released by developer alex09x, adds a `xref field` command that traces serialized field names to the code that writes their offsets, after an earlier query returned `not_found_under_limits` with no matching l…
Anthropic published a working theory of agent memory by Lance Martin describing "dreaming," an offline pass in which Claude rereads session transcripts to correct wrong notes, build linked wikis, and decide which facts l…