404 Media (RSS)
LLM News Today,only the few stories worth reading.
LLM News Today is a free daily feed of LLM news: we watch 878 sources, merge what they publish, and keep only the stories that change what you know about AI.
A diagram of the news pipeline: each line is one feed, lines gather into bundles where several feeds report the same story, and only a few bundles clear the gate into the daily digest.
Latest LLM News
The newest stories in LLM news, newest first — model releases, research, benchmarks and product updates from the last 30 days.
OpenAI (RSS)
Redwood Research: Blog (RSS)
Factory (RSS)
X: Rohan Paul (@rohanpaul_ai)
Ed Zitron / Where’s Your Ed At (RSS)
LLM News by Company & Model
The same LLM news, sorted by vendor and model family: every Claude, GPT, Gemini and Qwen release, and who shipped what.
OpenAI / ChatGPT
Everything OpenAI ships: the GPT line, ChatGPT, and the Sora product family.
View 64 picks →Anthropic / Claude
The Claude model line, Claude Code, and Anthropic’s safety research.
View 94 picks →Zhipu GLM
The GLM roadmap: open releases, coding ability, and reasoning gains.
View 3 picks →Google / Gemini
Gemini models, the Veo video family, and DeepMind research.
View 46 picks →OpenRouter
Model-routing platform news: new listings, usage stats, and developer strategy.
View 14 picks →NVIDIA
Chips and ecosystem: new GPUs, CUDA, and the robotics platform.
View 14 picks →Meta / Llama
Meta’s Llama family, its superintelligence lab, and compute spending.
View 26 picks →Cursor
Editor and code-intelligence news — the most watched line in AI coding tools.
View 4 picks →Hugging Face
Open-source community news: models, datasets, inference optimisation, and tooling.
View 5 picks →Microsoft / Copilot
Microsoft’s AI push: the Copilot stack, Azure AI infrastructure, and ecosystem.
View 12 picks →MiniMax
The M-series models, Hailuo speech, and multimodal work, including open-weight releases.
View 5 picks →Qwen / Alibaba
Alibaba’s Qwen releases and iterations, from flagship down to on-device models.
View 4 picks →Kimi / Moonshot AI
Moonshot’s K-series models, long-context work, and product moves.
View 6 picks →xAI / Grok
Musk’s xAI and Grok: model iterations, compute build-out, and the X platform.
View 6 picks →DeepSeek
Model releases, the open-source roadmap, and DeepSeek’s reasoning papers in one place.
View 11 picks →LLM News by Technology
LLM news by field — agents, multimodality, deployment, on-device models and safety — so you can follow one thread at a time.
Safety & Alignment
Jailbreaks and defences, model behaviour norms, and the case for capping frontier capability.
View 67 picks →Agents
Getting models to plan, call tools, and finish multi-step tasks on their own.
View 27 picks →AI Coding
Everything about AI writing code: assistants, vibe coding, and code-model benchmarks.
View 40 picks →Data & Training
The training side: data sourcing, synthetic data, cost curves, and method shifts.
View 8 picks →Open Source
Open weights, breakout community projects, and the open-versus-closed debate.
View 32 picks →Multimodal
Capability beyond text: vision, mixed image-and-text, and audio or video in and out.
View 22 picks →MCP & Tool Use
The most important protocol work outside the models themselves: the MCP ecosystem and tool calling.
View 11 picks →AI Video
Generation, understanding and post-production — where the limits are, and how production workflows are shifting.
View 14 picks →Reasoning
Where reasoning models stand: chain-of-thought, maths and logic benchmarks, and the arguments around them.
View 18 picks →Embodied AI
AI entering the physical world: humanoids, embodied models, and real-world manipulation.
View 3 picks →Speech & Audio
Speech synthesis, real-time voice, music generation, and audio understanding.
View 8 picks →Image Generation
The whole image pipeline: text-to-image models, editing, and the creative tool ecosystem.
View 5 picks →On-Device AI
AI that runs on phones, laptops and edge devices: small models and local inference.
View 12 picks →Deployment & Ops
The engineering of actually running models: inference optimisation, VRAM cost, and serving architecture.
View 25 picks →LLM News by Content Type
LLM news by format — research papers, benchmarks, tutorials and industry voices, kept apart from the day-to-day announcements.
Industry News
The business pulse of AI: funding, M&A, people moves, and the competitive map.
View 78 picks →Policy & Regulation
AI law and governance: legislation, regulation, and export-control frameworks.
View 7 picks →Industry Voices
What the key people think: founder interviews, researcher debates, and investor calls.
View 17 picks →Model Releases
New launches, open releases and iterations: vendor cadence and open-weight drops.
View 34 picks →Research Papers
AI papers worth your time: architecture, training methods, and capability measurement.
View 52 picks →Tutorials
Practical and ready to use: prompting, workflow building, and tool walkthroughs.
View 38 picks →Product Updates
Feature launches, redesigns and monetisation across AI products and apps.
View 60 picks →Trends
Changes already under way: usage shifts, capability limits, and social impact.
View 6 picks →Benchmarks
Who is actually ahead: benchmark results, evaluation disputes, and leaderboard movement.
View 20 picks →How LLM News Today Works
Four steps from a source we watch to the story on this page.
- 01
Crawl
878sources watched
Vendor blogs, release notes, research feeds and the X accounts that break model news first.
The list grows as the field does.
- 02
File
38topic desks
Every story is filed into the desks it belongs to — by company, by technology or by content type.
One story can sit on more than one desk.
- 03
Dedupe
399stories, last 30 days
The same story arriving from several sources is collapsed into one entry, inside a rolling 30-day window.
Newest first, always.
- 04
Publish
Dailypicks, no account needed
What survives ends up on this page and on the desk pages — free to read, and linked back to where it came from.
Every story links to its source.
LLM News FAQ
What the feed covers, where it comes from, and what it costs.
What is LLM News Today?
A free, daily feed of LLM news — model releases, research, benchmarks and product updates from across the field, cut down to the stories that actually change what you know.
How often is LLM news updated here?
Refreshed as the crawler reads its sources, and the home page always shows the last 30 days, newest first — so the top of the feed is the newest thing we have.
Which sources do you watch?
The crawler currently watches 878 sources, and the list moves as the field does: vendor blogs and release notes, research feeds, and the individual accounts that break news first.
How do you decide what makes the feed?
A story has to arrive with an English original headline — that is the one rule the feed enforces on the way in. Every story is then filed into a topic desk, and the same story coming in from several sources is collapsed into a single entry.
Do I need an account to read it?
No. There is no sign-up, no paywall and no app to install — open the page and read.
Can I follow one company, model or content type?
Yes. Every section on this page is also a page of its own — Anthropic, OpenAI, Benchmarks, Tutorials and the rest — so you can read one thread and skip everything else.