← sources

OVH announces Olala mini 7A1B

source: reddit (reddit:LocalLLaMA)published 2026-11-17 10:14ingested 2026-11-18 02:42

retained text

Olala mini is based on a 7-billion-parameter architecture (7B), with one billion active parameters spread across 256 experts per layer, and hybrid attention combining Differential-TPA and Mamba-3, in BF16 format. Only a fraction of the model is used for each request, making it faster and less memory-hungry. As a direct result, the model runs on a GPU with as little as 8 GB of memory, reaching 160 tokens per second on an RTX A4500, and even without a GPU, on a standard processor, at 50 tokens per second. Its context window currently reaches 32,000 tokens. Olala mini will be available as open source on Hugging Face, under the Apache 2.0 licence.

document fetched 2026-11-18 00:46 · canonical https://reddit.com/r/LocalLLaMA/comments/1xygtj4/ovh_announces_olala_mini_7a1b