A question picked from the 17 records added this week. We had read the move to open weights as narrowing on the license side, but this week two records moved the other way.
Week 43, 2026 · Oct 19–25 · updated every Monday
It is the first model from a new US lab. Company-reported scores were SWEBench Verified 81.6%, GPQA Diamond 90.6%, AIME 2026 97.7% and Terminal Bench v2.1 80.1% (on BF16), and it ships in three versions, BF16, FP8 and NVFP4, with the NVFP4 checkpoint at about 280GB
The authors reported that in the setting where the constraints have to be found by the model, even the best configuration averaged only 0.56 points and solved just 14 of the 95 tasks. Even in the 23 tasks where every necessary fact was obtained from the stakeholder, only 10 scored 0.8 or above, and they concluded that finding the unstated constraints is the bottleneck
In a company-reported sample (Sep 21–27, 2026; about 250,000 US conversations), politics, policy and society topics made up 5.75%, elections and voting 0.51%, conversations asking who to vote for 0.01%, and election-integrity policy violations 0.002%. In its own evaluation the political balance score for Opus 5.5, Sonnet 5.5 and Haiku 5.5 was 99%, but an outside reviewer pointed to a tendency to add caveats that were not asked for and to the risk of missing an operation broken into steps that each look harmless
A new lab released a model with 501 billion parameters in total under the loosest license there is, and a small forecasting model opened up in much the same way. Right up to last month, the higher a model ranked the more it turned to a license of its own, but this week the cases on the other side piled up. That said, every performance figure released is company-reported, and there are no outside results redoing them yet, so it is early to call a trend from one week.
In a new evaluation modelled on the work of actually handing results over to a customer, even the best setup finished only 14 of 95 tasks. Where they got stuck was not ability but the part where they had to dig out, on their own, the conditions nobody had told them. Another study out the same week said that bundling several agents together raises the score but pushes the cost up three or four times, and as much as seven.
Laptops with unified memory raised to 128GB have started selling, and one forecasting model runs on an ordinary processor alone, with no graphics card. A bundle also came out that handles chat, judging and dictation in one place while leaving memory as it was. Apart from the race on performance, moves to change where things get used gathered in the same week.
The same date, in other years.
1,295 records and 106 notes in English so far. Tap a bar to open that year. 87% of the 1,295 records rest on a primary source.
Start with a five-minute lesson a day, then go deeper with the AI History Module and the Part Reviews.
One dot can only multiply and add. Stack them in layers, and out comes judgment.
The same record, rewritten for three levels. It's revised every week as new records come in.
When a part ends, it's summed up again at three depths. No tests.
New weekly records and each day's lesson go into the feed. Paste the address into any reader app.
/en/feed.xmlA language model does one thing very fast: it picks the next piece of a sentence. This site adds one more piece every week.