dnjfr retweeted
📢 Meet Qwen4-Max, your new AI teammate for work, creativity, and discovery!
Qwen4-Max is built on a new MoE architecture with native vision and a 1M-token context window.
Highlights:
- Outperforms Qwen3.8-Max with up to ~50% fewer tokens, while average cost per task drops over 60% in our internal evaluation.
- Advances real-world productivity through greater proficiency with professional tools, stronger long-horizon execution, and higher-quality deliverables.
- Handles real tasks, including 3D modeling, game development, manufacturing, finance, law, healthcare, and frontier scientific discovery.
- $1.5/M tokens for input, $4.5/M tokens for output, and $0.075/M tokens for implicit caching — all lower than Qwen3.8-Max, with implicit caching down by 70%.
The API is live on QwenCloud. You can also try Qwen4-Max on Qwen Studio, QwenWork and Qoder.
Start building today! 🚀
- Blog: qwen.ai/blog?id=qwen4-max
- API: qwencloud.com/models/qwen4-m…
- Token Plan: qwencloud.com/pricing/token-…
More to come! The open weights of smaller Qwen4 models will be released. Stay tuned. 👀
"Ceci est une œuvre de fiction. Toute ressemblance avec des personnes existantes ou ayant existé, ou avec des événements réels, serait purement fortuite." 🙃
営業部長「Outlookが使えないんだが」
情シス「Microsoft側で障害が発生しています」
営業部長「こんな忙しい時に。いつ直るんだよ?」
情シス「復旧時刻は未定です」
営業部長「はぁ?Microsoftには連絡したのか?」
情シス「公式の障害情報を確認しています」
営業部長「いや、電話しろよ」
情シス「……電話ですか?」
営業部長「そうだ。直接話した方が早いだろ」
情シス「世界規模の障害なので、現在対応中かと」
営業部長「じゃあ責任者を出してもらえ」
情シス「……Microsoftの責任者ですか?」
営業部長「そうだ。こっちは金払ってるんだろ?」
情シス「世界中の企業が払っています」
営業部長「うちは長年使ってるんだ。優先してもらえ」
情シス「……」
営業部長「とにかく電話して、何時に直るか聞いてくれ。それか、うちだけ先に直してもらえるよう交渉しろ」
情シス「世界規模の障害なんですが」
営業部長「そこを何とかするのが情シスの仕事だろ?」
情シス「……分かりました。確認します」
営業部長「最初からそうすればいいんだよ」
情シスは、静かにMicrosoftのサポートページを開きました。
そこには「復旧時刻未定」と書かれていました。
情シスは、しばらくその画面を見つめていました。
その間にも、営業部長からの催促電話は鳴りやみませんでした。
Microsoftの復旧時刻は、誰にも分かりませんでした。
ただ一つ、分かっていることがありました。
営業部長からの次の電話は、必ず来るということです。
本当に怖いのは、Microsoftの障害じゃない。
月額料金を払えば、情シスがMicrosoft本社を動かせると思われていることだった。
2
55
👀
Introducing Apex Neural Rendering (ANR), our cross-platform Super Resolution, Ray Reconstruction, and Frame Generation technology, delivering up to 2.4× faster performance and higher quality rendering than NVIDIA DLSS.
ANR rebuilds neural rendering around a new architecture that outperforms current state of the art models.
Works on all GPU vendors.
- NVIDIA: GeForce RTX 20, 30, 40, and 50 series, plus GTX 16 series
- AMD: Radeon RX 5000, 6000, 7000, and 9000 series, plus Steam Deck
- Intel: Arc A and B series
Frame Generation available now:
apexcompute.com/neural-rende…
Our Super Resolution and Ray Reconstruction models are currently in training. Coming soon!
24
dnjfr retweeted
Bring your ideas to life with Zeus, a new type of GPU:
- Expandable memory: 32/64/128 GB soldered + 2x/4x SO-DIMM slots for up to 384 GB memory!
- Massive improvements to path tracing performance!
- 400 GbE QSFP-DD port!
- Built-in high-performance RISC-V CPU cores capable of running Linux!
- DevKits 2026, Mass Production 2027
And we used the good old 8-pin PCIe power connector that is known to not melt!
Plus que jamais, on a besoin d'une nouvelle alternative crédible à Nvidia et AMD sur les GPU grands publics
NVIDIA announces CEO transition. Read the release: nvidianews.nvidia.com/news/n…
1
61
I just solved "Robotics Engineering" education with Fable 5.2 forever..
no need to spend $500+ on a starter kit to learn what a GPIO pin does, just visit my interactive site:
• 8 real boards in 3D: Arduino Uno, Raspberry Pi 4...
• hyper-realistic graphics with real-life actions
• build part by part: place chips, connectors and headers, then solder them in
• wire real circuits: blink an LED, read a sensor, drive a motor
Fable 5.2 generated every 3D model in code, no ready-made 3D files or photos
try it free, no sign-up ↓
robotics-lab-app.vercel.app
dnjfr retweeted
Qwen4 is coming.
Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities.
Built for real-world tasks — from complex coding and knowledge work to front-end builds and professional software workflows.
Qwen4-Max is now in preview on Alibaba’s Token Plan. You can also get early access to the new model on QwenWork, Qoder, Qwen APP, and QwenStudio.
Put it to the test on a real problem. Show us what you build. 👀
Token Plan: qwencloud.com/pricing/token-…
Promis, après celui là on arrête de sortir des modèles encore plus puissants et on fait une pause dans la course à l'IA 🙃
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
2
142
dnjfr retweeted
En Chine🇨🇳, des enseignants peuvent corriger les copies manuscrites d'une classe en 5 mn grâce à une IA!!
L'enseignant insère une pile de devoirs,
l'appareil les scanne un par un, analyse les réponses, attribue des notes et imprime directement ses corrections
Via @_3emeOeil
Qu'est ce que c'est pénible la nouvelle popup Reddit qui t'invite à te connecter, il n'y a même pas de bouton pour la fermer, on doit réactualiser la page 💀
1
51
C'est intéressant de voir ce que chaque LLM a sous le capot
Catching up on the 10 (!) open LLM architecture releases from the past month. Quite a lot happened in October.
I guess we can call it "Open October"!
1. Kolibri-1 (78B): A notable model from Germany. Features classic Grouped Query Attention (GQA) and Sliding Window Attention (SWA) attention module.
2. Ling 3.1 Flash (560B): Uses Kimi Delta Attention and the eighth uses gated Multi-head Latent Attention (MLA) inspired by DeepSeek.
3. Muse Spark Open (816B): Besides Musse Glimmer, the first major open weight release from Meta (since Llama). Uses classic GQA + SWA but adds a LatentMoE.
4. Nemotron 3.5 Super VL (120B): Combines Mamba-2 layers with just eight attention layers. Also LatentMoE
5. Dory1.0 (4B): A looped transformer design.
6. Reflection Beam (501B): GQA+SWA architecture.
7. Recursive Flint (40B): Note that "Recursive" is the company name; the architecture is a fairly standard MoE with GQA and SWA.
8. MiniMax M3.1-Flash (428B): Uses their own MiniMax Sparse Attention module.
9. LFM3 (1.7B): Hybrid architecture that mixes short convolutions and attention.
10. Mistral Large 4 (1T) combines MLA with sparse attention.
Will probably cover/discuss things in more depth in the coming weeks!
1
50
dnjfr retweeted
𝗔𝗴𝗲𝗻𝘁 == 𝗟𝗟𝗠 + 𝗛𝗮𝗿𝗻𝗲𝘀𝘀
An agent is just an LLM and some additional scaffolding aka harness. This harness can be split into:
1. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁: your prompt, system prompt, additional instructions (e.g. in CLAUDE .md), possibly other context like files, …, conversation history
2. 𝗧𝗼𝗼𝗹 𝗰𝗮𝗹𝗹 𝗳𝗼𝗿𝗺𝗮𝘁: agreed format in which the LLM specifies the tool it wants to run and with which arguments (e.g. read_file, edit_file, bash, search, …)
3. 𝗚𝘂𝗮𝗿𝗱𝗿𝗮𝗶𝗹𝘀: rules and permissions that are enforced by the harness (some actions are always ok, some are never ok, some require confirmation)
4. 𝗥𝘂𝗻𝘁𝗶𝗺𝗲: code that actually executes the tool call and passes the results back to the LLM as additional context. This runtime can also contain MCP clients to talk to external tools (databases, …).
5. 𝗖𝗵𝗲𝗰𝗸𝘀: when the LLM thinks it is done, the harness can run configured checks (aka hooks), for example run the test suite or type checker. If any of these checks fail, this information is passed to the LLM and it can try again. Only when the checks pass will the harness indicate success in the interface.
The interface can be a terminal app (such as Claude Code), GUI app (e.g. Codex), etc. In principle an interface can use multiple harnesses and a harness can be used from multiple interfaces.
So what is actually the LLM's job? To generate text! Nothing more. The LLM cannot actually do anything on your computer, it just generates text, some of which may trigger actions in the harness. This is the core loop of an agent: the LLM wants to call a tool, the harness executes it and passes the results back to the LLM.
This means when you're switching between coding agents you're not just using a different LLM, but a different harness!




























