LFM2.5-8B-A1B doubles its tokenizer vocab and cuts on-device decoding time up to 3.7x
maximelabonne · x · 2026-07-22
A report and blog post describe how LFM2.5-8B-A1B upgraded its tokenizer vocabulary from 65K to 128K to better handle languages that had been split too finely.
The result was much shorter token sequences and faster on-device decoding:
- Thai: 4.0× fewer tokens
- Vietnamese: 2.6× fewer tokens
- Hindi: 2.4× fewer tokens
- Estimated 2.2–3.7× faster per-character decoding on device for these languages
The post frames this as a recipe for upgrading a pretrained model's tokenizer in place.
More from Infra
- PyTorch details 2.12x faster on-device streaming ASR on Arm SME2 via ExecuTorch — PyTorch · 2026-10-27
- Revenue per GPU can rise as token prices fall, argues Gavin Baker — GavinSBaker · 2026-10-27
- Anthropic S-1: only $19B total compute spend since inception, overbuild warning — bookwormengr · 2026-10-27
- CoreWeave, You.com and NVIDIA lift BrowseComp 22.9% in 3-week RL post-training sprint — RichardSocher · 2026-10-27
- Grid queues make power architecture an AI-factory decision; Exowatt and Redwood offer fixes — ingliguori · 2026-10-27
- Two CVSS 9.9 Wasmtime flaws allow sandbox escape, one ARM64-only since 2020 — basedjensen · 2026-10-27