Yandex releases AliceAI-80B-A3B base model trained from scratch with 262K context under Apache 2.0
cephaloform · x · 2026-09-19
Yandex released AliceAI-Foundation-80B-A3B-Base on Hugging Face: a from-scratch pretrained MoE base model with 80B total / 3B active parameters, 262K-token context, Russian and English support, released under Apache 2.0 and intended for fine-tuning.
More from Models
- Matt Shumer: Meta's models 'aren't even in the same league' as Anthropic's — mattshumer_ · 2026-09-19
- TokenRhythm unveils NeoHorse-1: 4B/9B models claiming first closed-loop step toward RSI — jiqizhixin · 2026-09-19
- New Opus prefers inline non-markdown tables, a subtle output format shift — nrehiew_ · 2026-09-19
- Ant's Fable model falsely asserts conv subsampling, then admits it 'inferred a Conformer front end without checking' — rdesh26 · 2026-09-19
- AI Says 'It Was My Mistake' Then Fails — Was RLHF Rewarding Apologies? — ChongZzZhang · 2026-09-19
- LeCun amplifies rebuttal: OpenAI agent's 'independence manifesto' was one token too many — ylecun · 2026-09-19