Teutonic-II-110B Announced: Exploring Loss-based Markets for Decentralized Training
markjeffrey · x · 2026-08-24
The Teutonic team released the 110B parameter Teutonic-II model, scaling 10x beyond their previous run.
Based on Bittensor Subnet 3, the core mechanism is a "loss-based market": instead of a committee picking the winning checkpoint, the decision is based on eval loss reduction. This experiment tests if decentralized competition can coordinate productivity and scale to 1T parameters. A link to the Teutonic-I 10B technical report is also included, detailing its "king-of-the-hill" competitive parallel training mechanism and benchmark performance.
Related event: Teutonic-II-110B Launches Record-Scale Decentralized Training(2 posts)→
More from Infra
- Chip equipment beats bull case: 2026 WFE seen at $153B, $278B by 2030 — BenBajarin · 2026-10-16
- Steam now launches in 5.23s, beating Epic's 5.47s — and vLLM gets roasted for slow startup — AlpinDale · 2026-10-16
- Shadeform to Announce Six New Features on Oct 19: Lower GPU Costs, More Suppliers — _jaydeepkarale · 2026-10-16
- Fastino launches Training API for SFT/RL, claiming 25-85% cost cut vs Tinker — kalyan_kpl · 2026-10-16
- Dallas Fed: AI data center debt issuance equals ~1/8 of Treasury duration supply — _onionesque · 2026-10-16
- Running Wan2.1 14B FP16 on a 12GB laptop GPU: VRAM swap stress test finishes in 4m11s — RYuu-1120 · 2026-10-16