Chutes releases Parallax 8B Theta model
Chutes released the Parallax 8B Theta model after training it across distributed consumer GPUs. Founder Jon Durbin presented the project at Exploit Summit as a blueprint for cheaper, more resilient AI infrastructure.
Parallax is a compelling infrastructure demonstration, but its decentralized training claims still need broader independent replication before challenging centralized clusters.
- –Trained across 240 RTX 5090 GPUs spanning 13 countries
- –Extends Chutes from decentralized inference into decentralized model training
- –Targets lower-cost, permissionless compute using ordinary consumer hardware
- –The model’s local-inference angle strengthens the privacy and resilience case
- –Reported performance is promising, but production-scale reliability remains unproven
DISCOVERED
55m ago
2026-10-01
PUBLISHED
1h ago
2026-10-01
RELEVANCE
AUTHOR
AlphaGapTAO