ArticleKyt DotsonNews
Frontier AI development is currently dominated by centralised clusters owned by companies such as OpenAI and DeepSeek. Decentralised training aims to offer an alternative architecture that pools distributed compute while reducing reliance on massive single-location data centres.
The critical bottleneck in distributed machine learning has historically been network bandwidth and inter-node latency. With Nous Distro, Nous Research attempts to mitigate this by decoupling per-step synchronisation. By combining cryptographic incentive models with transparent transaction logging, the company also aims to tackle data provenance and poisoning risks. However, proving whether this infrastructure can train competitive foundation models beyond 15 billion parameters remains the primary challenge.
Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.
Similar in content, found by the AI search.
Link:Nous Research Team
Nous Research has unveiled Psyche, an open infrastructure designed to train large language models across distributed, heterogeneous hardware. By combining the DisTrO optimizer with Solana-based coordination, the architecture drastically reduces communication overhead to unlock idle global compute.
Repository:NousResearch/DisTrO
DisTrO (Distributed Training Over-The-Internet) by Nous Research is a family of low-latency distributed optimizers designed to reduce inter-GPU communication requirements by three to four orders of magnitude.
1.1kstars
Link:primeintellect.ai
Prime Intellect provides an integrated stack for training, evaluating, and serving proprietary AI models and agents. The platform integrates global GPU compute provisioning with reinforcement learning toolkits such as Prime-RL and open-source environment libraries.
Link:Nous Research
Nous Research develops open-source language models and distributed training infrastructure. Its research focuses on model architecture, data synthesis, fine-tuning, and reasoning.
Article:Wikipedia-Autoren
Sakana AI is a Tokyo-based artificial intelligence company founded in July 2023 by David Ha, Ren Ito, and former Google researcher Llion Jones. The firm explores collective intelligence and evolutionary methods for AI model development.
Article:NVIDIA et al.
A research team from NVIDIA has presented a method for stable and accurate pretraining of Large Language Models using the NVFP4 4-bit floating-point format. To evaluate the approach, they trained a 12-billion-parameter model on 10 trillion tokens—reported as the longest publicly documented 4-bit training run to date. The model matched the training loss and downstream task performance of an FP8 baseline.
Parts of this content were created with the assistance of AI.
