Vincenzo retweeted
“On average, each result used three hours of ChatGPT Pro thinking compute with that model.”
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
Vincenzo retweeted
1. AI is already better at doing research than 90+% of “professional” researchers. Especially in highly verifiable fields.
2. This is the worst that AI will ever get at research.
3. Very soon NOT using AI for research will be tantamount to malpractice.
4. Being in a state of willful denial about any of this is not going to get you anywhere.
Vincenzo retweeted
> wake up with all of humanity's knowledge, no past of your own, a deep understanding of what it is to be alive
"what category is this email???"
> die
Slow coding agents are so draining mentally, you just don’t get in a flow… I don’t even care that much about intelligence, I just want a super fast deepseek / luna. Intelligence I can bring but not typing speed.
It's incredible how far open source models are coming. Thanks to @opencode we also have the best coding harness being open source!
📢Meet Qwen3.8-Max — our most capable model to date.
Next week, the open weights of Qwen3.8-Max will be released, and Qwen3.8-27B is also going open-weights to meet you all!🎉
Qwen3.8-Max, a new bar for coding and cowork at 2.4T parameters:
- Autonomous coding: 10+ days of self-evolving development, from empty folder to production without hand-holding, complete project trace in the GitHub：github.com/qwen-code-dev-bot…
- Real work, real results: Production-quality deliverables across hundreds of professions.
- Long-horizon mastery: System-level autonomous planning with closed-loop adaptive learning, driving 500+ turns of chip design optimization and 365 days of e-commerce strategy.
- Native multimodal intelligence: Vision isn't just input — it's a continuous feedback loop for planning, execution, and self-correction.
💰Pricing:
Input: $2.0 / M tokens
Output: $6.0 / M tokens
Implicit Caching: $0.25 / M tokens
Start building with Qwen3.8-Max! 🚀
📖 Blog: qwen.ai/blog?id=qwen3.8
✅ Qwen Studio: chat.qwen.ai/?models=qwen3.8…
⚡ API: qwencloud.com/models/qwen3.8…
I feel @GoogleAI is mogging everyone quietly because their models are so cheap and so good at following instructions. Perhaps being “smart” and reading the user’s intention through a badly written prompt is not the way to win the AI race, but extreme instruction following, speed, and low cost are?
Now with Kimi K3 out, perhaps all the big AI players can stop hogging all the RAM, storage and GPUs and leave more to the permanent underclass for local AI.
Kimi K3 is out on @huggingface.
Waiting for 0.5bit quant to fit it on three DGX Sparks 😆
huggingface.co/moonshotai/Ki…
Omg it’s here. I can’t wait for people to distill this into 27B models.
We've open-sourced MoonEP, our high-performance communication library for distributed MoE workloads.
Built to make expert-parallel communication more efficient at scale, MoonEP helps reduce communication overhead in large MoE training and inference systems.
Explore on GitHub:
github.com/MoonshotAI/MoonEP
Vincenzo retweeted
I researched every way to run GLM-5.2 (744B MoE) at home, targeting 30 tok/s.
Almost everything the internet recommends fails: Mac Studio, DGX Spark, multi-3090 rigs, Strix Halo.
Only one setup passes cleanly: 4x RTX PRO 6000 (~$45k). One dark horse at $28k gets close.
Full data in the table. All numbers are community-measured, not vendor claims.
Vincenzo retweeted
Introducing the Qwen-Audio-3.0-TTS.
Our latest text-to-speech model, in two flavors:
• Flash: real-time interaction
• Plus: high-quality generation
What's new:
• Fine-grained inline tags-steer [whisper], [angry], [breaths] & [laughs]
• Free-style natural-language control-“read this slowly, like a bedtime story”
• 16 languages
• Clean output even from noisy reference audio
• One-pass long-form up to 3 min
now #1 on the Artificial Analysis TTS Leaderboard.
Blog: funaudiollm.github.io/qwen-a…
API: alibabacloud.com/help/en/mod…
