After 10B beat 72B, TAO's Teutonic-II launches 110B decentralized training run
bittingthembits · x · 2026-08-24
Bittensor subnet SN3 Teutonic announced Teutonic-II-110B, a frontier permissionless training of a 110B LLM — 10x larger than its previous model and 50% larger than any prior decentralized run.
The previous Teutonic-I 10B context: over 70 days and 2,163 duels, only 203 challengers (9.3%) dethroned the incumbent via a strict statistical test. The final model averaged 62.28% across 11 benchmarks, beating Covenant 72B (57.55%) and Quasar-Preview 18B (59.78%), crushing INTELLECT-1 10B and Psyche 40B, and leading 8 of 11 individual benchmarks.
Mechanics: no gradient averaging or shared runs — one king checkpoint at a time, trained king-of-the-hill style by uncoordinated strangers, yet outscoring a 72B model.
Related event: Teutonic-II-110B Launches Record-Scale Decentralized Training(2 posts)→
More from Models
- Cornell and Amazon revisit the 16x pretraining compute gap: diffusion LMs need only 3-8x — LucaAmb · 2026-10-16
- Image Decision Benchmark JevImageBench hits v0.3.0 with reshuffled Top 5 — airesearch12 · 2026-10-16
- Forecast: Meta's Muse Spark Open May Become the Best Non-Chinese Open-Weight LLM on Agentic Benchmarks — zsakib_ · 2026-10-16
- Claude Opus 5.5 effort levels tested: low reasoning iterates 9x faster than max — tom_doerr · 2026-10-16
- Matt Turck podcast unpacks Jev: TypeSafe's machine-native AI and its RLHF critique — mattturck · 2026-10-16
- LoOPSD: looping a model's own layers as teacher lifts Qwen3-1.7B GSM8K from 42% to 80% — burny_tech · 2026-10-16