Last updated on January 2nd, 2026 at 06:35 pm
Quick Breakdown
- QVAC Genesis II adds 107 billion tokens to the original dataset, covering new fields like chemistry and machine learning.
- The Option-Level Reasoning method analyses all answer choices in a question to improve AI reasoning accuracy.
- Dataset released openly under CC-BY-NC 4.0 license to support researchers outside proprietary systems.
Tether Data advances open AI training
Tether Data announced QVAC Genesis II on December 22, 2025, growing the world’s largest public synthetic educational dataset for AI pre-training to 148 billion tokens. This release adds 107 billion tokens to QVAC Genesis I, now spanning 19 domains, including chemistry, computer science, statistics, machine learning, astronomy, geography, econometrics, and electrical engineering. College-level physics content was regenerated using improved methods to improve quality.
Meet Genesis V2. We’ve introduced “Option-Level Reasoning”—a method that trains models to understand why distractors are incorrect, not just pick the right answer.
📈 The Result: ~30% avg accuracy (competitors ~12%). ✅ Reliability: 99.4% valid, clear answers.
Read the paper &… pic.twitter.com/gTXEvC2jr5
— QVAC (@QVAC_tether) December 22, 2025
The expansion uses Option-Level Reasoning, a technique that breaks down every multiple-choice option, correct or incorrect, to teach causality and address misconceptions. This pairs with the prior Failure Analysis approach from Genesis I, creating data that prioritizes clear explanations over raw volume. Models trained on this data show higher reasoning scores and produce unambiguous outputs, per independent tests.
Tether CEO Paolo Ardoino stated the focus shifts from fluency to structured understanding in AI training.
“Intelligence should be built on understanding why something is true, not just predicting what sounds right,”
he said. The dataset supports local, decentralized AI to reduce reliance on centralized clouds services.
Broader impact on decentralized intelligence
Notably, Tether led an $8M investment in Speed to build Lightning-based, USDT-settled payment infrastructure for global merchants, expanding the use of stablecoin payments.
QVAC Genesis II aligns with Tether Data’s goal of peer-to-peer systems for secure data sharing without intermediaries. Available on Hugging Face under the Creative Commons Attribution-NonCommercial 4.0 license, it aids academics and developers worldwide. A technical paper details the methods on the QVAC research blog.
This move counters industry trends of scraping vast volumes of text, instead building reasoning-focused data. Tether Data, part of the Tether ecosystem, drives innovation in privacy and efficiency for digital networks. QVAC’s mission, “Local AI. Infinite Intelligence. No Compromise targets device-based AI for communities.
The release comes amid rising demand for high-quality open datasets amid the dominance of proprietary models. Independent evaluations confirm Genesis II outperforms earlier synthetic data in clarity and decision-making.
If you would like to read more articles like this, visit DeFi Planet and follow us on Twitter, LinkedIn, Facebook, Instagram, and CoinMarketCap Community.
Take control of your crypto portfolio with MARKETS PRO, DeFi Planet’s suite of analytics tools.”
