CompaniesA kids' coding school now sells a $199 AI course.
· 3 min read
ExplainedThe AI that invents a world as you walk through it.
· 5 min readTrending
AI API pricing tracker: What every major model costs per million tokens
What every major AI model costs through its API, per million input and output…
Nvidia reportedly eyes Firmus stake weeks after the AI cloud pulled its IPO
According to The Information, Nvidia is negotiating a multi-hundred-million-dollar investment in Firmus, the Australian…
Advertisement
Latest News
Pretraining data filtering: How AI labs remove dangerous knowledge before a model learns it
Data filtering removes risky documents from an AI model's training set, so it never learns them. Here is how it works, what research shows,…
Moonshot trimmed Kimi K3.1 Pro’s hacking skill before opening the weights
Moonshot says it stripped cyber-related data from training for Kimi K3.1 Pro, cutting its score on an internal hacking test from 90.5%…
Block’s Dessa talks and listens at once, and runs on a laptop
Block says its new Dessa model learned from human conversation alone, interrupts and backchannels like a person, and runs locally on a…
Prompt caching: How AI APIs reuse repeated text to cut cost and delay
Prompt caching lets an AI service skip text it has already read. How it works, how Anthropic and OpenAI price it, where…
Bridge rounds: How companies fund the gap before a big raise or an IPO
A bridge round is short-term funding that carries a company from one big raise to the next, or to an IPO. Here…
Trusted access: How AI labs unlock sensitive biology models for vetted users only
Trusted access lets AI labs give vetted users a less restricted model while the public gets stricter safeguards. Here is how it…
Featured
New AI models this week: Kimi K3.1 Pro and DeepSeek V4.1 Flash lead
Moonshot's Kimi K3.1 Pro leads a week of three new API models and open releases from…
A 501B open model you can own: Reflection’s Beam ships under Apache 2.0
Reflection has published Beam, a 501B-parameter mixture-of-experts model for coding and agents, with open weights under…
Midjourney AI image generator: How it works, what it costs and the free alternatives
Midjourney costs from $10 a month with no general free tier. This guide covers how it…
A new Hugging Face Space shows which Hub models run on AWS Trainium
Hugging Face and AWS Neuron published a searchable Space and a public dataset that report, model…
Hugging Face now hosts 2,000+ free simulation-ready 3D assets for robots and AI factories
Nvidia's SimReady Central Library gathers more than 2,000 free OpenUSD assets with collision, mass and friction…
xAI’s Grok Bot teammate cleared 204 support tickets in 10 minutes
xAI describes two Grok Bot teammates: IncidentWatch flags spikes in similar tickets, and IncidentSweep answers the…
OpenAI
OpenAI nears a $30B bridge round at a $1.25T valuation
- OpenAI cut ChatGPT’s read latency 12x, from 120 ms to 10 ms
- OpenAI asks Newsom to toughen the California AI law it once opposed
- OpenAI’s GPT-6 models reach Google Cloud, with no prices yet
Google’s Chrysalis agent fixed a simulated outage in 1.8 seconds, with humans signing off
- Google Cloud’s Vera Rubin A5X instances arrive next month, Europe to follow
- Gemini 4 Argon hit nearly 25T tokens in 72 hours, 3x Gemini 3
- Gemini in your car can now read your Gmail, Photos and Maps
Anthropic
Claude wrote 180 million lines of proofs to check 123 open-source libraries
- Anthropic hands Agent Skills to a neutral foundation, with 280 million installs
- Claude on Google Cloud now has an Anthropic-run option, with the same price
- Claude API gets its first new version since 2023, and old SDK code will break
Advertisement
Top categories
AI Tools & Guides
What the tools do, what they cost, and how to start
Mistral Le Chat is now Vibe: The models, plans and prices of Mistral’s AI assistant
Mistral's Le Chat was renamed Vibe in May 2026. Here is what the assistant does, which models sit…
Perplexity AI vs ChatGPT for research: Sources, search and pricing compared
Perplexity is built around cited web answers, ChatGPT is a broader assistant with search added; here is how…
Gemini AI Studio: What is free, what costs money and how it differs from the Gemini app
Google AI Studio is Google's free web workspace for testing Gemini models and getting an API key. This…
Microsoft Copilot Studio: What you can build, what it costs and how to start
Copilot Studio is Microsoft's low-code studio for agents and workflows. This guide covers what it builds, the credit-based…
Advertisement
More News
Explore more stories from Aiholics News
Tencent’s Hy-Nomas world model, now in Diffusers, runs on one GPU
Tencent has put its Hy-Nomas interactive video world model on Hugging Face in Diffusers format, with Apache-2.0 weights and a…
AI film studio Utopai reportedly raises $60M at a $1B valuation
The Information reports that Utopai, an AI film studio with more than 20 projects in the works, raised $60M at…
Model distillation: How a small AI learns from a bigger one, and why it is disputed
Distillation trains a small, cheap AI model on the answers of a bigger one. It is a standard technique, and…
AI tools for small business: What to use for writing, support, accounting and marketing
A job-by-job guide to AI tools for small businesses: Claude and Gemini for writing, Intercom for support, QuickBooks and Xero…
Kimi K3.1 Pro is on OpenRouter, with 1M context and weights due by Nov 4
Moonshot's Kimi K3.1 Pro is now callable through OpenRouter. Moonshot says it has a 1M-token window, always-on reasoning, and open…
Nvidia’s video search model finds events in footage, but only for research
Nvidia's fine-tuned InternVideo2-CLIP ranks surveillance-style clips by text query and gains 8.78 points of mAP over its base model, but…
FCC lets Intel keep upgrading foreign-made robots in its US chip fabs
The FCC's engineering office granted Intel a one-year waiver for hardware, software and firmware changes to automated guided vehicles in…
Nvidia’s 4GB Nemotron Bolt model is for local assistants, game characters and robots
Nvidia says its newest small open model, Nemotron 3.5 Bolt E4B 4GB, is available now for local assistants, interactive characters…
Advertisement