Nous Research introduces Token Superposition Training for efficient LLM pretraining
The method uses bag-of-tokens during initial training to speed up wall-clock time by 2-3× at matched FLOPs.
See the latest news and media coverage for Nous Research. We track all announcements, press releases, and industry mentions in real time, all in one place.
Open source AI research organization
nousresearch.comLast updated
In short: Nous Research launched its Hermes 4 and 4.3 models, debuted the Hermes AI Agent, and secured a $50 million investment at a $1 billion valuation.
The method uses bag-of-tokens during initial training to speed up wall-clock time by 2-3× at matched FLOPs.
It achieves 67.87% Pass@1 on LiveCodeBench v6, trained with reinforcement learning on 24k problems.
After proving distributed training at scale, they will run ablations and train Hermes 4 on Seed-OSS-36B model.
The Forge API enhances LLMs with reasoning capabilities, while Nous Chat offers a platform for Hermes 3. Both are now available.
Nous Research's open-source AI agent is the first with a built-in learning loop—it creates skills fr...
Nous Research launches Hermes 4 open-source AI models that outperform ChatGPT on math benchmarks with uncensored responses and hybrid reasoning capabilities.
Leo Schwartz is a reporter at Fortune covering fintech, crypto, venture capital, and financial regulation. Training AI models requires a massive amount of computing power...
Ultimately, the DisTrO method could open the door to many more people being able to train massively powerful AI models.
Track Nous Research and your other target companies to get real-time alerts and weekly summaries delivered straight to your inbox.
Browse news for competitors to Nous Research and other trending companies.