The Frontier - SiliconNoon (Page 17)

The Frontier

GPT-6 Astra outscores every specialist at antibody prediction

The rule that AI-for-science requires dedicated models just broke. OpenAI's GPT-6 Astra topped a blind benchmark for predicting whether antibodies can become drugs — beating purpose-built systems without any domain-specific training. GPT-6 Astra scored 37.98 on the DDD Benchmark's antibody developability module, outperforming every specialized model in a blind evaluation built by Insilico Medicine. The benchmark integrates six sets of antibody experimental data and tests a question that has bur

The Frontier

Xiaomi open-sources CocktailASR-1 to pick one voice out of a crowd

Xiaomi just gave away one of the harder pieces of the voice stack. On September 11 the company released and open-sourced Xiaomi-CocktailASR-1, an end-to-end speech recognition model built to solve the cocktail party problem — transcribing one specific person's speech from an audio mix where several people are talking over each other, without a separation stage in front of the model. The design is unusually direct. You feed the model a short reference clip of the target speaker (one to four seco

The Frontier

Princeton's Mengdi Wang: LLMs haven't made a real scientific discovery yet

The models that cracked competition math still can't find a new particle — and Princeton's Mengdi Wang thinks she knows why. Large models learn "the most likely" answer, and science lives in the long tail. That was the core of a keynote from Mengdi Wang, director of Princeton's Center for AI Innovation, at the Inclusion·Bund Conference in Shanghai on September 10: the training method behind today's LLMs has conquered math and coding but, in her words, still hasn't produced a genuinely new disco

The Frontier

Cognition's SWE-2 closes in on the frontier — at a quarter of the price

Coding agents stopped being a two-horse race today. Cognition has launched SWE-2, the model behind Devin, and by its own published numbers it lands closer to the frontier than any release in the company's history — while undercutting the models it trails on cost. On the FrontierCode 1.1 Main and DeepSWE 1.1 coding benchmarks, Cognition says SWE-2 beats its own SWE-1.7 and xAI's Grok 4.6 on both score and cost, matches GPT-5.6 Sol and Anthropic's Fable 5/5.1 at a fraction of their price, and com

The Frontier

GigaChat 3.5 Reasoning lands under MIT — and beats DeepSeek on math

Sber's GigaChat team has open-weighted a 432-billion-parameter reasoning model under an MIT licence, and the numbers put it within touching distance of DeepSeek V4 Flash. GigaChat 3.5 Reasoning runs 28 billion active parameters, trains natively in FP8, and mixes multi-head latent attention with GatedDeltaNet linear-attention layers — a hybrid aimed squarely at long-context cost — with a 262K-token window and three speculative-decoding heads bolted on. The release includes FP8 weights for inferen

The Frontier

Amap's world model builds a 3D city in 10 minutes on one GPU

Alibaba's mapping arm used its Street Stars anniversary to make a bigger claim: it now has a world model that generates explorable 3D space, not just renders it. Two smaller stories round out the hour — a computer-vision benchmark that found audio evidence beats extra parameters, and a fresh report that reasoning state can still be replayed into another model. Alibaba's Amap released ABot-Earth 0.7, a 3D-native world model that turns a single satellite image or a text prompt into a kilometer-sc

The Frontier

Envision's Tianji is now the best weather model nobody in the West talks about

Three stories from Shanghai's Bund Summit and the Chinese feeds this morning: an energy company's weather model just took the crown on the field's most-cited benchmark, Alipay's assistant started booking real travel, and a Nobel laureate put a number on what AI is worth. Envision's Tianji weather model entered Google's WeatherBench 2 for the first time this week and landed in the top tier alongside Google's own WeatherNext 2 — beating it on wind speed and precipitation, the two variables the en

The Frontier

DeepSeek V4.1 Flash ships — and retires V4 Pro next Monday

DeepSeek's mid-morning release isn't just a new checkpoint. It's a new architecture, an open-weights drop, and a forced migration for anyone still calling its flagship — all in one announcement. DeepSeek released V4.1 Flash on Wednesday, the first model in a new architecture family and the first with native visual understanding. The headline number is 552 billion parameters, but the point is how few of them run: a new causal encoder–decoder design activates just 8 billion parameters while readi

The Frontier

Ant's AI reads gastric tumors to predict who relapses after surgery

Two stories this morning are about the same underlying bet: that AI earns its place by being right about something narrow and measurable. One is a gastric cancer model validated on thousands of real patients; the other is a regulator deciding that some uses of AI are too socially expensive to leave alone. Ant Group's medical AI lab and Hebei Medical University's Fourth Hospital have published three models that predict what surgeons currently guess at — who will have serious complications after

The Frontier

Deep Dive — Math's credit system was built for humans. AI just broke it.

A Millennium Prize problem appears to have fallen, and the interesting part is not the proof. It is that the institutions mathematics uses to decide who discovered something — priority, publication, peer review, the two-year wait before a prize — all assume a human pace and a human author. OpenAI's Navier–Stokes result went from rumor to Lean-verified in five days. Nothing in the rulebook is calibrated for that, and the fight that erupted is what happens when a discovery outruns the machinery me

The Frontier

Two tokens of Claude's thinking made Kimi K3 sound like Claude

A distillation-detection paper went looking for stolen reasoning traces and found something stranger: feed Moonshot's open-weight model a few words of Claude's thinking and it starts writing like Claude. Plus DeepSeek's next Flash model is about to swallow its own Pro tier, and Shopify cut a 6,000-token agent prompt down to 1,500. A paper on stealing chain-of-thought turned up an awkward signal about Kimi K3. "Stealing Reasoning Traces from Proprietary LLM APIs" — a paper co-authored by former

The Frontier

OpenAI splits its image model: Flare for speed, Sunburst for control

Two releases worth your attention this morning: OpenAI has turned its image model into a two-tier product line, and the former head of Huawei's Pangu model has shipped his first model out of a startup built on watching other models work. OpenAI released GPT-Image-2.5 on September 8 — but the real change is that it shipped as two models instead of one. Flare is the fast default, which OpenAI says produces better images than GPT-Image-2 at up to half the latency; Sunburst is the slower, precision

See all