Tether AI Research achieves 99% valid-answer rate with Genesis III model

Tether's AI research division QVAC released Genesis III, a synthetic STEM dataset of 191.43 billion tokens, on September 23, 2026, and reported that a 1.7 billion-parameter model trained on part of the data reached a 99.45% valid-answer rate on MMLU STEM benchmarks, with the accompanying paper accepted at COLM 2026. The dataset spans roughly 159.6 million documents across 19 STEM domains in three difficulty tiers, split into 108.67 billion tokens of Option-Level reasoning data and 82.76 billion tokens of Failure Analysis data. Against the Cosmopedia-v2 baseline, Genesis III-trained models improved by 28.57 points on ARC-Easy, 21.35 points on ARC-Challenge, and 15.03 points on MMLU STEM; the dataset carries a CC-BY-NC 4.0 license and the model an Apache 2.0 license.

Tether AI Research achieves 99% valid-answer rate with Genesis III model The stablecoin giant's AI research arm released a massive synthetic STEM dataset that pushes small language models into surprisingly competitive territory Tether https://cryptobriefing.com/markets/tether/ ’s AI research division, operating under the name QVAC, has released Genesis III, a synthetic STEM dataset containing 191.43 billion tokens. A 1.7 billion-parameter model trained on a portion of that dataset hit a 99.45% valid-answer rate on MMLU STEM benchmarks, a result strong enough to earn the accompanying research paper acceptance at COLM 2026. What Genesis III actually is The dataset spans approximately 159.6 million documents across 19 different STEM domains, organized into three difficulty tiers: high school, college, and professional. QVAC split the corpus into two distinct training methodologies. The first, called Option-Level OL reasoning, trains models by focusing on correct answers and the logic behind them, accounting for 108.67 billion tokens. The second, Failure Analysis FA , does the opposite: it trains models on errors and why they’re wrong, contributing another 82.76 billion tokens. The model trained on the OL split alone produced the headline 99.45% valid-answer rate. But the full-corpus models told an even more compelling story in comparative benchmarks. Against Cosmopedia-v2, a well-known baseline dataset, Genesis III-trained models improved by 28.57 points on ARC-Easy, 21.35 points on ARC-Challenge, and 15.03 points on MMLU STEM. AI, tech, and the markets they move—in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. The Genesis trilogy Genesis III is the culmination of a series that started with Genesis I, released on October 24, 2025, with 40.91 billion tokens. Genesis II followed on December 22, 2025, expanding coverage to 19 domains and adding 107.46 billion tokens. Genesis III, released on September 23, 2026, completes the trilogy. Tether has released the dataset under a CC-BY-NC 4.0 license, meaning anyone can use it for non-commercial research and education. The model itself carries an Apache 2.0 license, which is more permissive and allows commercial use. Why the stablecoin company is doing AI research Tether has been steadily diversifying beyond stablecoins. The stated goal is to democratize access to AI education tools, enabling local tutoring and technical assistance without relying on centralized cloud services. Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy https://cryptobriefing.com/editorial-policy/ .