minishlab ships blink CPU embedding models; benchmarks show 3.6x faster single queries
tomaarsen · x · 2026-11-18
minishlab released the blink embedding collection (blink-8M at 7.68M params and multilingual blink-multilingual-64M), designed to run on CPU via model2vec, model2vec-rs, or sentence-transformers.
sentence-transformers maintainer tomaarsen ran a quick CPU check (i7-13700K, short texts):
- Single queries: model2vec was 3.6-3.8x faster than Sentence Transformers across both models;
- With batchsize=1024: Sentence Transformers was 2.2x faster;
- The Rust model2vec-rs implementation likely beats both.
A practical lightweight option for embedding retrieval on CPU.
More from Models
- After DeepSeek's rate limits: a 4-step guide to the official paid desktop path — sujingshen · 2026-11-18
- blink-8M tops seven static embedding models with 55.35 MTEB and 46.70 BEIR nDCG@10 — tomaarsen · 2026-11-18
- minishlab's Blink: 8M-parameter static embeddings run on CPU without a transformer pass — tomaarsen · 2026-11-18
- PhilBench quizzes 97 LLMs with the 2020 PhilPapers Survey of 1,785 philosophers — pwlot · 2026-11-18
- GPT-6.2 Sol tops browser-use leaderboard at 82%, 35% cheaper and 2x faster than predecessor — charliermarsh · 2026-11-18
- Decision models beat LLMs at structured output: Liquid AI d1 runs 2.7-3.6x faster than LFM3 — mervenoyann · 2026-11-18