Sentence Transformers maintainer: stop using all-MiniLM-L6-v2, it's leaving performance on the table
tomaarsen · x · 2026-09-16
Sentence Transformers author tomaarsen reacted to all-MiniLM-L6-v2 trending on Hugging Face with a blunt note: it's a 2021 model and you're leaving performance on the table. In follow-ups he suggests newer, smaller alternatives like IBM's granite-embedding-30m-english and Snowflake's snowflake-arctic-embed-s, points to the 26,221 sentence-transformers-compatible models on HF, and advises starting from the MTEB leaderboard but ultimately evaluating on your own data before picking one.
More from Models
- xAI Launches Grok 4.6 With Unified Multi-Modal API, Sub-200ms Latency — Jasonio · 2026-09-16
- AI Detectors Crumble Against Fine-Tuned Models: Pangram Falls to 3% Accuracy, GPTZero to 0% — Dr_WhoDo · 2026-09-16
- Mozilla Report: China-US AI Model Capability Gap Narrows to 4.4 Months — External_Mood4719 · 2026-09-16
- Claude free plan quietly shows Opus 5 option, limited to just 3 messages — AssociationShoddy785 · 2026-09-16
- Dev recalls fighting for structured output in GPT-3 days as dynamic inputs now parse in <200ms — blixt · 2026-09-16
- llama.cpp Merges hc Ops for Qwen4exp, Time to Re-benchmark Qwen Flash — jacek2023 · 2026-09-16