Qabbel

ERNIE 4.5 21B-A3B Thinking vs ERNIE 4.5 VL 28B-A3B

Qiegħed żewġ mudelli sa erbgħa ma' xulxin fuq l-oqsma li jiddeċiedu deployment: kuntest, prezz, reġjun, liċenzja, lingwi, hosting, u x'hu verifikat.

ERNIE 4.5 21B-A3B ThinkingERNIE 4.5 VL 28B-A3B
Fornitur BaiduBaidu
Parametri 21.8B29.4B
Tieqa tal-kuntest Mhux verifikat: il-fornitur ma ddokumentax din id-dikjarazzjoni.Mhux verifikat: il-fornitur ma ddokumentax din id-dikjarazzjoni.
Liċenzja apache-2.0apache-2.0
Modalitajiet texttext, vision
Data tar-rilaxx 08 Set 202528 Ġun 2025
Lingwi Qawwi: en
Adekwat: zh
Qawwi: en
Adekwat: zh
Piżijiet miftuħa ivaiva
Kompiti Raġunament, Spiża baxxaViżjoni, Raġunament

Fejn żewġ mudelli jvarjaw, din it-tabella turi d-differenza u tieqaf. LLM EU ma tippubblikax badge ta' rebbieħ mingħajr punteġġ dokumentat għal dik il-kompitu, u figura nieqsa tibqa' nieqsa minflok ma ssir żero.

Meta tużah

ERNIE 4.5 21B-A3B Thinking

Saħħiet

  • 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
  • the model page lists en
  • instruction-tuned checkpoint with a chat template in the repository

Limiti

  • 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 21.8B parameters, self-hosting needs a single accelerator with enough memory for the weights

Kompiti

Raġunament, Spiża baxxa

ERNIE 4.5 VL 28B-A3B

Saħħiet

  • 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
  • the model page lists en

Limiti

  • 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 29.4B parameters, self-hosting needs a single accelerator with enough memory for the weights
  • a base checkpoint: it answers nothing useful without fine-tuning or a task head
  • vision checkpoints are heavier to serve than their parameter count alone suggests, because the image encoder and pre-processing run alongside the language model

Kompiti

Viżjoni, Raġunament

Kompiti: Raġunament

Qabbel mudelli

Qabbel mudelli →