LaBSE En Ru vs Rubert Tiny2

Wins 0 of 0 benchmarks
Wins 0 of 0 benchmarks
No benchmarks in this slice.

Specs

SpecLaBSE En RuRubert Tiny2
Context1K2K
WeightsOpenOpen
ReleasedJun 2021Oct 2021
Dimensions—312
Max input1K tokens2K tokens
Languagesen, ruru
Memory (4-bit)0.1 GB0 GB
Licence—MIT
Commercial use—Allowed

What breaks when you switch

LaBSE En Ru → Rubert Tiny2
Re-embed your whole corpus›

Vectors from different models can't be mixed.

Languages: loses en
Rubert Tiny2 → LaBSE En Ru
Re-embed your whole corpus›

Vectors from different models can't be mixed.

Max input: 2K tokens → 1K tokens
Memory (4-bit): 0 GB → 0.1 GB