Colqwen3.5 4.5B v3 vs ColUltraVBERT

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

Specs

SpecColqwen3.5 4.5B v3ColUltraVBERT
Context262K33K
WeightsOpenOpen
ReleasedMar 2026Oct 2026
Max input262K tokens33K tokens
Languagesen, es, fr +2en, es, fr +3
Memory (4-bit)2.8 GB0.2 GB
LicenceApache-2.0MIT
Commercial useAllowedAllowed

What breaks when you switch

Colqwen3.5 4.5B v3 → ColUltraVBERT
Re-embed your whole corpus›

Vectors from different models can't be mixed.

Max input: 262K tokens → 33K tokens
ColUltraVBERT → Colqwen3.5 4.5B v3
Re-embed your whole corpus›

Vectors from different models can't be mixed.

Languages: loses pt, it
Memory (4-bit): 0.2 GB → 2.8 GB