Xiaohongshu open-sources dots3-note, its IMO perfect-score LLM

Share
Xiaohongshu open-sources dots3-note, its IMO perfect-score LLM

China's AI week keeps delivering: a social-commerce giant just opened the weights of the first model family to ace the International Mathematical Olympiad, and the country's top foundry is cashing in on AI demand like never before.

Xiaohongshu's dots lab open-sourced dots3-note preview, the first public release from the dots3 series — the same family as dots-note-3.0, which last month became the first AI system to score an official perfect 42/42 at the International Mathematical Olympiad. The 280-billion-parameter MoE (16 billion active) carries a 512K context window and multimodal understanding across text, vision, and audio. Rather than chasing another math leaderboard, dots3-note preview is aimed at long-horizon agent tasks — the messy, open-ended jobs like travel planning, wedding prep, or running a small store that stretch across hours or days and have no single right answer. That positioning tracks where the agent market is heading: tasks that take a human more than an hour to finish.

The lab says the model matches or beats models several times its size on reasoning and agent benchmarks, with vision performance that stands out in its weight class — which would make the IMO gold look less like a party trick and more like a genuine reasoning core. The full dots3 lineup will span note, jazz, and aria tiers, so this is the beachhead of a family rather than a one-off. It is also the latest sign that China's biggest consumer platforms are treating open weights as table stakes, not a differentiator.


SMIC, China's largest foundry, raised wafer prices and posted record revenue above $3 billion as AI demand tightens chip supply. Co-CEO Zhao Haijun said on the earnings call that the company negotiated price increases with customers in the first quarter and will charge more for wafers processed in the third quarter. "Future wafer starts are far exceeding our previous expectations," he said, with the company weighing more capacity as AI-related orders keep beating forecasts and drive what he expects to be a strong second half. The pricing power is notable for a foundry that historically competed on cost: AI demand is now so concentrated that even China's most sanctioned chipmaker can name its price.

What to watch: the rest of the dots3 family (jazz, aria) — and whether SMIC's hikes flow through to AI hardware prices later this year.

An IMO-perfect model aimed at wedding planning instead of math olympiads — is that the right use of a reasoning breakthrough? Tell us in the comments.

Read more

Byteification turns Qwen3 and Llama 3 into byte-level models

Byteification turns Qwen3 and Llama 3 into byte-level models

The tokenizer is the last piece of the language-model pipeline that was never learned end to end — and a Nature paper out this week shows it can be swapped out for less than one percent of a pretraining budget. Researchers have retrofitted four open subword language models into true byte-level models using less than 1% of a typical pretraining budget, and the result — published 7 October in Nature — closes the performance gap that has kept byte-level models a niche curiosity for nearly a decad

OpenAI's dots go mobile and can start your Codex work

OpenAI's dots go mobile and can start your Codex work

One product update and two benchmark papers this hour, all circling the same question: how much of the work can you actually hand over, and what happens to the part you didn't? OpenAI's dots — the always-on agents that get their own cloud computer — can now be created and configured entirely from the ChatGPT app on iOS and Android, and a dot can start Codex work or pick up an existing Codex thread on your behalf. The October 9 release, billed as the first of a fresh run of updates for dots, mo

Coding agents as a hospital: Cockroach's five-month results

Coding agents as a hospital: Cockroach's five-month results

The agentic-coding debate keeps being framed as speed versus trust. Cockroach Labs just published the most detailed trust-first ledger yet — five months of numbers, including the parts that didn't work. Cockroach Labs ran its codebase like a teaching hospital for five months and says the pipeline merged well over a million lines of code with seven total reverts. In the setup the company calls MOLT Sinai, GitHub issues are patients, merging is discharge, and the humans in charge are Chiefs of M

Today in AI — October 10, 2026

Today in AI — October 10, 2026

A day where the boring machinery took center stage: proof checkers, order books, warehouses and the humble text message — plus fresh arguments about who gets to declare any of it safe. Models & Research * Mathematicians get a reliability primer for the Lean Theorem Prover. A guest post by Thomas Hales on Terence Tao's blog walks through what Lean actually guarantees — and devotes itself to the "Summer of Soundness Bugs," the string of kernel bugs found this summer that let false proofs thr