Usporedi
Qwen3.5 2B vs Qwen3.5 9B
Stavite dva do četiri modela jedan pored drugoga na poljima koja odlučuju o implementaciji: kontekst, cijena, regija, licenca, jezici, hosting i što je verificirano.
| Qwen3.5 2B | Qwen3.5 9B | |
|---|---|---|
| Pružatelj | Alibaba Qwen | Alibaba Qwen |
| Parametri | 2.3B | 9.7B |
| Prozor konteksta | Nije provjereno: pružatelj nije dokumentirao ovu tvrdnju. | Nije provjereno: pružatelj nije dokumentirao ovu tvrdnju. |
| Licenca | apache-2.0 | apache-2.0 |
| Modaliteti | text, vision | text, vision |
| Datum izdanja | 28. velj 2026. | 27. velj 2026. |
| Jezici | Nije provjereno: pružatelj nije dokumentirao ovu tvrdnju. | Nije provjereno: pružatelj nije dokumentirao ovu tvrdnju. |
| Otvoreni utezi | da | da |
| Zadaci | Vizija, Niska cijena, Višejezična podrška | Vizija, Višejezična podrška, Zaključivanje |
Kada se dva modela razlikuju, ova tablica pokazuje razliku i tu staje. LLM EU ne objavljuje značku pobjednika bez dokumentiranog rezultata za taj zadatak, a brojka koja nedostaje ostaje nedostajuća umjesto da postane nula.
Kada ga koristiti
Qwen3.5 2B
Prednosti
- open weights are published on Hugging Face and can be downloaded without a vendor agreement
- Apache-2.0 licence, so commercial self-hosting is permitted without a separate agreement
- at 2.3B parameters it fits on a single accelerator
Ograničenja
- catalogue only — LLM EU does not run this model, and nothing on this card implies an LLM EU endpoint
- no evaluation has been run by LLM EU
- the published configuration records no context window, so the card carries none
- at 2.3B parameters, the weights fit on a single modest accelerator
- vision checkpoints are heavier to serve than their parameter count alone suggests, because the image encoder and pre-processing run alongside the language model
Zadaci
Qwen3.5 9B
Prednosti
- open weights are published on Hugging Face and can be downloaded without a vendor agreement
- Apache-2.0 licence, so commercial self-hosting is permitted without a separate agreement
- at 9.7B parameters it fits on a single accelerator
Ograničenja
- catalogue only — LLM EU does not run this model, and nothing on this card implies an LLM EU endpoint
- no evaluation has been run by LLM EU
- the published configuration records no context window, so the card carries none
- at 9.7B parameters, the weights fit on a single modest accelerator
- vision checkpoints are heavier to serve than their parameter count alone suggests, because the image encoder and pre-processing run alongside the language model
Zadaci
Zadaci: Vizija, Višejezična podrška