Inside the Model Factory — Eiso Kant, Poolside AI
Quick Answer
Poolside AI's new Laguna S 2.1 model, with 118 billion parameters, outperforms larger models from competitors by nearly tenfold.
Quick Take
The company emphasizes rapid development, achieving model launches in just eight weeks, and advocates for open research in AI.
Key Points
- Laguna S 2.1 features 118 billion parameters and 8 billion activated per token.
- Poolside AI can conduct 10,000–20,000 experiments monthly with fewer than 70 researchers.
- The company raised $500 million while promoting open-source AI and global research.
- Model launches reduced from six months to just five to eight weeks.
- Poolside emphasizes the importance of persistence and verification over raw intelligence.
DeepSignal Analysis
What happened
Poolside AI has launched its Laguna S 2.1 model, which features 118 billion parameters and claims to outperform larger models from competitors by nearly tenfold. The company has developed a Model Factory that allows for rapid model launches within eight weeks. Their recent technical report has received positive feedback for its detail.
Key evidence
- Laguna S 2.1 is a Mixture-of-Experts model with 118 billion total parameters and 8 billion activated per token, designed for high performance.
- Poolside AI's Model Factory enables the company to conduct 10,000 to 20,000 experiments monthly, significantly speeding up the model development process.
- The company raised $500 million to support its operations, indicating strong investor confidence in its approach to AI development.
Why it matters
The emergence of Poolside AI with its Laguna S 2.1 model highlights the competitive landscape in AI, particularly in the context of open versus closed model ownership. Their rapid development cycle and emphasis on open research could influence industry standards and practices. Additionally, the significant funding received may enable further advancements in AI technology and research.
Source Excerpt
Poolside's co-CEO on how his small team of top researchers built a model factory capable of training Laguna S - a 118B MOE beating Thinky's ~1T open weights model... and this is just the beginning.
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from Latent Space
See more →![[AINews] SpaceXAI launches Grok 4.5, first Opus-class model post Cursor acquisition](https://substackcdn.com/image/fetch/$s_!8D6O!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fpbs.substack.com%2Fmedia%2FHMuQw2BXUAAJaQd.png)
[AINews] SpaceXAI launches Grok 4.5, first Opus-class model post Cursor acquisition
SpaceXAI has launched Grok 4.5, a new coding-focused model that is 3x larger than Grok 4.3, priced at $2 per million input tokens. Positioned as an Opus-class model, it aims for efficiency and speed, outperforming competitors like GPT-5.6 and Opus 4.8 in cost-effectiveness.
