EmbeddingGemma 2 sets new Pareto frontier across code, image, audio benchmarks
tomaarsen · x · 2026-10-28
Google's EmbeddingGemma 2 technical report shows the model establishing a new Pareto frontier across code (MTEB), image/video/visual documents (MMEB-v2), audio (MAEB), and image (MIEB-lite) benchmarks, beating models like UltraBERT-150M, CLAP, and SigLIP at small parameter counts.
Sentence Transformers maintainer tomaarsen called the Pareto charts impressive and plans to read the report for nuggets he can port into Sentence Transformers — he's also listed in the report's acknowledgements.
Related event: Google Open-Sources EmbeddingGemma 2, a 744M Omni-Modal Embedding Model(7 posts)→
More from Models
- Google ships Gemini 4 Carbon two weeks after Argon, claims SOTA on cyber and coding — tulseedoshi · 2026-10-29
- Early Take on Gemini 4 Carbon: Lots of Thinking, Good Model Overall — SamuelAlbanie · 2026-10-29
- Gemini 4 Carbon tops AA Index v5, beats Opus 5.5 and GPT-6 Astra on most benchmarks — Yuchenj_UW · 2026-10-29
- Gemini 4 Carbon Keeps $2/$10 Pricing, Scores 59 on AA v5.0 — _philschmid · 2026-10-29
- Reddit thread: picking third-party LLM inference providers, GLM Flash cited as best value — inwardPersecution · 2026-10-29
- Reddit reposts Gemini 4 Carbon announcement with benchmark table — Marimo188 · 2026-10-29