Sakana AI announces its own pretrained LLM series Sakana Sawara, first model due spring 2027
SakanaAILabs · x · 2026-11-22
At its Tokyo symposium "From World Models to Real-World AI" with 400+ attendees, Sakana AI revealed it is developing its own pretrained LLM series, tentatively named Sakana Sawara, with chief scientist Takuya Akiba announcing that the first small model, Sawara Small, is slated for release in spring 2027.
Jürgen Schmidhuber, who joined as Chief Scientific Advisor in September, gave a keynote covering his foundational contributions (early Transformers, pretraining/distillation, world models, adversarial artificial curiosity), and CEO David Ha joined him for a fireside chat on how the 2018 World Models paper came about and what it means to be a scientist in the AI era. Sessions on open-ended autonomous discovery and Physical AI round out the event; full recordings with Japanese subtitles are on YouTube.
More from Companies & People
- NYT's AI projects editor writes every draft on a 1950s Olivetti typewriter, never with AI — dylfreed · 2026-11-22
- Schmidhuber traces neural nets to Gauss; Sakana AI lays out Japan AI sovereignty pitch — SakanaAILabs · 2026-11-22
- Sakana AI publishes Tokyo symposium videos, unveils self-pretrained Sawara model line — SakanaAILabs · 2026-11-22
- OpenAI and Anthropic week 47: GPT-6.2 Astra/Sol launch, Anthropic's Nasdaq debut, watermarking — btibor91 · 2026-11-22
- Second GEO Conference wraps up: brands optimizing to become AI's answer — vista8 · 2026-11-22
- Dario Amodei reportedly used an air-gapped computer and printer at OpenAI — ScottNover · 2026-11-22