NVIDIA's Nemotron team reports gold-threshold scores at IMO 2026 and IOI 2026 using one open base model family, and it published the checkpoints, data, prompts and proofs. A plain-language look at what the result shows, what it does not, and why reproducibility may matter more than the medals.
Liquid AI has released two open-weight models that return yes/no, multiple-choice and scored answers in a single forward pass instead of generating tokens, with vendor-reported latency as low as 16 milliseconds on an NVIDIA Jetson edge board. The benchmarks are the company's own; independent verification is absent.
NVIDIA's newly released Kumo Tabular reads labeled spreadsheets and predicts outcomes for new rows in a single pass, with no training or feature engineering. Here is what is verified, what is vendor-run, and why it could change who can build predictive tools.
H Company released Holo4 as open weights in 27B dense and 35B-A3B Mixture of Experts sizes on September 28, 2026, scoring 61.7 percent on OSWorld 2.0 for the 27B model against 81.8 percent for Opus 5.5, publishing 7,366 agent trajectories for public inspection, and shipping a smaller Holotron4 Nano. Cost comparisons are self-reported, and the weights carry a non-commercial license.
Ai2's AstaBrief 8B turns a research question and retrieved excerpts into a cited report in one pass. The weights and training data are now public, cutting pipeline report time from 178.5 to 51.1 seconds, but Ai2's own caveats about 2025-era benchmarks and unmeasured claim overstating deserve equal billing.
OpenAI's Global Affairs team published "Building standards for the next phase of AI" on September 21, 2026, urging international technical standards for frontier AI and recursive self-improvement through CAISI and the AI safety institutes network, while stating the standards would not be licenses, mandatory prerelease review, or approval requirements.