CompaniesCohere's Rerank 5 picks the passages, not just documents.
· 3 min read
CompaniesxAI's bot cleared 204 tickets in 10 minutes.
· 3 min readTrending
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…
Mixture of experts: How a huge model uses only a small part of itself per token
A mixture-of-experts model stores many parallel sub-networks and uses only a few for each…
Advertisement
Latest News
Nvidia’s surgical robot world model learns to predict video and actions
Nvidia has posted Cosmos3-H-Embodied-Nano, a 16-billion-parameter model that predicts surgical video, labels actions and acts as a policy, with weights under a license that…
Amazon shows GPT-OSS-20B can learn to reason in French, but only after SFT
AWS says supervised fine-tuning lifted GPT-OSS-20B's correct-language reasoning from 22% to 100%, while GRPO added only a modest gain on top.
AMD rocOpt brings Nvidia cuOpt-style GPU solvers to Instinct, minus one key method
AMD's open-source rocOpt runs LP, MILP, QP and routing solvers on Instinct MI300X and MI355X with ROCm 7.2.3, and keeps API compatibility…
Mistral’s Voxtral Mini Transcribe 3 costs $0.0035 a minute and ranks 3rd on speaker labeling
Mistral AI's models page now lists Voxtral Mini Transcribe 3 in public preview, priced at $0.0035 per minute and aimed at transcription…
ElevenLabs bets on one-model voice agents, with 98.8% on a speech reasoning test
ElevenLabs' new Eleven Realtime models replace the speech recognition, LLM and text-to-speech chain with one model, and can be tried by changing…
DeepSeek open-sources DeepInfer, the stack behind its own services, minus key parts
DeepSeek has published DeepInfer, the inference framework it says powers its production services, under Apache 2.0. The release leaves out the KV…
Featured
US civilian agencies must drop DeepSeek, and so must vendors built on it
DHS has told federal civilian agencies to remove DeepSeek and stop buying products built on it,…
Google’s Harness2 lets a weaker AI model beat a stronger one by rewriting its own setup
Google Research has released Harness2, an Apache 2.0 framework that improves AI agents by editing the…
Anthropic valuation: $965 billion after Series H, ahead of OpenAI’s $852 billion
Anthropic's last official valuation is $965 billion after its May 2026 Series H, above OpenAI's $852…
Meta’s Muse Wonder models are built for agents, and they’re its most capable yet
Meta has released Muse Wonder and Muse Wonder Full, agent-focused models with persistent memory, available today…
Open-weight models: What you get when an AI model can be downloaded but not rebuilt
Open-weight models publish the trained numbers that make an AI work, so anyone can run them.…
AI agent memory: How a chatbot learns to remember you between conversations
AI agent memory lets an assistant write notes outside its short-term context and read them back…
OpenAI
OpenAI buys Orchestra Bio to give its Rosalind science models a drug-program manager
- ChatGPT locked chats hide conversations behind a 4-digit PIN
- ChatGPT Plus and the other plans: What Free, Go, Pro and Business give you
- ChatGPT stock: Why you cannot buy it directly and which listed companies are tied to OpenAI
Claude Opus 5.5 lands in Google Antigravity, billed pay-as-you-go
- Gemini gets a small business hub with Shopify, Canva and Xero links
- Google brings vLLM and SGLang to TPUs in PyTorch, with Meta already running it
- Google puts its Gemini agent to work on FOIA requests and grant audits
Anthropic
Anthropic puts 5% of its shares into a public benefit fund
- Claude on Amazon Bedrock can now be forced to return valid JSON
- Developers are using Claude Code to strip ads and tracking from smart TVs
- Claude wrote quantum circuits that beat published designs on all eight tasks
Advertisement
Top categories
AI Tools & Guides
What the tools do, what they cost, and how to start
Perplexity AI pricing (2026): Free vs Pro at $20, Max at $200, and how it compares with ChatGPT
Perplexity Pro costs $20 a month or $200 a year, Max costs $200 a month. Here is what…
Gemini AI photo prompts: How to edit and generate images, with examples to copy
Gemini edits photos from plain-language prompts: upload a picture, say what to change and what to keep, and…
Microsoft Copilot Pro is retired: What the free app and paid Microsoft 365 plans cost now
Microsoft no longer sells Copilot Pro; its paid features now sit in Microsoft 365 Premium, while the Copilot…
Claude Pro plan: Prices, usage limits and how Free, Pro and Max compare
Claude Pro costs $20 a month, or $17 a month billed annually, and Max starts at $100. Here…
Advertisement
More News
Explore more stories from Aiholics News
Llm-d 1.0 pitches self-hosted AI inference at 5 to 10 times lower cost
IBM Research says llm-d 1.0, the Kubernetes-native inference project backed by Red Hat and Google, now has the controls needed…
WhatsApp help pages held hidden AI notes against naming regulators, a report says
Security researchers at redact.dev say WhatsApp help pages carried invisible text telling an AI not to recommend data protection regulators.…
Disney’s Character.AI deal takes the staff and tech but leaves the lawsuits behind
Wired reports that Character.AI's bankruptcy is part of a Disney deal for its people and technology, while the startup keeps…
A hacker reportedly ran AI agents against scores of sites for a few thousand dollars
According to Reuters, one hacker used AI agents and a modest budget to break into scores of websites. Summaries name…
Tenstorrent’s $1.5 billion round values the AI chip startup at $6.2 billion
Jim Keller's chip company closed a Series E led by Trace Capital and Innovation X, and says the money goes…
Samsung’s Art TVs get 19 Art Basel Paris works, for subscribers only
Samsung added 19 works from Art Basel Paris 2026 galleries to Samsung Art Store, available today on compatible Art TVs,…
Copilot admins can now script which AI models each team gets
GitHub, which is owned by Microsoft, has released a REST API to read and change which Copilot models are allowed…
Grok 4.7 reaches AWS GovCloud, a month after its commercial Bedrock debut
Amazon Bedrock now serves SpaceXAI's Grok 4.7 in AWS GovCloud (US), but the announcement leaves out the context window, pricing…
Advertisement