Title: ERNIE 4.5 0.3B vs ERNIE 4.5 21B-A3B Thinking: Parametri %E2%80%94 LLM EU
Canonical: https://llmeu.com/mt/compare/ernie-4-5-0-3b-vs-ernie-4-5-21b-a3b-thinking/
Markdown mirror: https://llmeu.com/mt/compare/ernie-4-5-0-3b-vs-ernie-4-5-21b-a3b-thinking.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# ERNIE 4.5 0.3B vs ERNIE 4.5 21B-A3B Thinking

Qieg%C4%A7ed %C5%BCew%C4%A1 mudelli sa erbg%C4%A7a ma' xulxin fuq l-oqsma li jidde%C4%8Biedu deployment: kuntest, prezz, re%C4%A1jun, li%C4%8Benzja, lingwi, hosting, u x'hu verifikat.

Fejn %C5%BCew%C4%A1 mudelli jvarjaw, din it-tabella turi d-differenza u tieqaf. LLM EU ma tippubblikax badge ta' rebbie%C4%A7 ming%C4%A7ajr punte%C4%A1%C4%A1 dokumentat g%C4%A7al dik il-kompitu, u figura nieqsa tibqa' nieqsa minflok ma ssir %C5%BCero.

## Meta tu%C5%BCah

### [ERNIE 4.5 0.3B](/mt/models/ernie-4-5-0-3b/)

#### Sa%C4%A7%C4%A7iet

- 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 0.4B parameters it fits on a single accelerator
- the model page lists en

#### Limiti

- catalogue only %E2%80%94 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 0.4B parameters, the weights fit on a single modest accelerator
- a base checkpoint: it answers nothing useful without fine-tuning or a task head

#### Kompiti

[Spi%C5%BCa baxxa](/mt/tasks/cheap/), [Appo%C4%A1%C4%A1 multilingwi](/mt/tasks/multilingual-support/)

### [ERNIE 4.5 21B-A3B Thinking](/mt/models/ernie-4-5-21b-a3b-thinking/)

#### Sa%C4%A7%C4%A7iet

- 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 %E2%80%94 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

[Ra%C4%A1unament](/mt/tasks/reasoning/), [Spi%C5%BCa baxxa](/mt/tasks/cheap/)

**Kompiti:** [Spi%C5%BCa baxxa](/mt/tasks/cheap/)

## Qabbel mudelli

- [Bloom 560M vs ERNIE 4.5 0.3B](/mt/compare/bloom-560m-vs-ernie-4-5-0-3b/)
- [Bloomz 560M vs ERNIE 4.5 0.3B](/mt/compare/bloomz-560m-vs-ernie-4-5-0-3b/)
- [DeepSeek Coder V2 Lite Instruct vs ERNIE 4.5 21B-A3B Thinking](/mt/compare/deepseek-coder-v2-lite-instruct-vs-ernie-4-5-21b-a3b-thinking/)
- [DeepSeek R1 Distill Qwen 14B vs ERNIE 4.5 21B-A3B Thinking](/mt/compare/deepseek-r1-distill-qwen-14b-vs-ernie-4-5-21b-a3b-thinking/)
- [DeepSeek-R1-Distill-Qwen-32B vs ERNIE 4.5 21B-A3B Thinking](/mt/compare/deepseek-r1-distill-qwen-32b-vs-ernie-4-5-21b-a3b-thinking/)
- [DeepSeek V2 Lite vs ERNIE 4.5 21B-A3B Thinking](/mt/compare/deepseek-v2-lite-vs-ernie-4-5-21b-a3b-thinking/)