Title: Phi 3 Mini 128K Instruct vs Phi-4: Parametrit, Lisenssi %E2%80%94 LLM EU
Canonical: https://llmeu.com/fi/compare/phi-3-mini-128k-instruct-vs-phi-4/
Markdown mirror: https://llmeu.com/fi/compare/phi-3-mini-128k-instruct-vs-phi-4.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# Phi 3 Mini 128K Instruct vs Phi-4

Aseta kaksi tai nelj%C3%A4 mallia vierekk%C3%A4in kentille, jotka ratkaisevat k%C3%A4ytt%C3%B6%C3%B6noton: konteksti, hinta, alue, lisenssi, kielet, is%C3%A4nn%C3%B6inti ja mik%C3%A4 on vahvistettu.

Kun kaksi mallia eroavat, t%C3%A4m%C3%A4 taulukko n%C3%A4ytt%C3%A4%C3%A4 eron eik%C3%A4 jatka pidemm%C3%A4lle. LLM EU ei julkaise voittajamerkki%C3%A4 ilman dokumentoitua pisteytyst%C3%A4 kyseiselle teht%C3%A4v%C3%A4lle, ja puuttuva luku pysyy puuttuvana sen sijaan, ett%C3%A4 siit%C3%A4 tulisi nolla.

## Milloin k%C3%A4ytt%C3%A4%C3%A4

### [Phi 3 Mini 128K Instruct](/fi/models/phi-3-mini-128k-instruct/)

#### Vahvuudet

- open weights are published on Hugging Face and can be downloaded without a vendor agreement
- mit licence, so commercial self-hosting is permitted without a separate agreement
- at 3.8B parameters it fits on a single accelerator
- the model page lists en
- instruction-tuned checkpoint with a chat template in the repository

#### Rajoitukset

- 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 3.8B parameters, the weights fit on a single modest accelerator

#### Teht%C3%A4v%C3%A4t

[P%C3%A4%C3%A4ttely](/fi/tasks/reasoning/), [Yhteenveto](/fi/tasks/summarization/), [Haku (RAG)](/fi/tasks/rag/)

### [Phi-4](/fi/models/phi-4/)

#### Vahvuudet

- strong MMLU and maths scores at 14B parameters
- MIT licence without user thresholds
- fits a single 40GB accelerator at BF16

#### Rajoitukset

- 16K-token context window
- declared English-only; EU-language performance is not documented
- training leans on synthetic data, which narrows domain robustness

#### Teht%C3%A4v%C3%A4t

[P%C3%A4%C3%A4ttely](/fi/tasks/reasoning/), [Yhteenveto](/fi/tasks/summarization/), [Edullinen](/fi/tasks/cheap/), [Haku (RAG)](/fi/tasks/rag/)

**Teht%C3%A4v%C3%A4t:** [P%C3%A4%C3%A4ttely](/fi/tasks/reasoning/), [Yhteenveto](/fi/tasks/summarization/), [Haku (RAG)](/fi/tasks/rag/)

## Vertaa malleja

- [Aya 23 8B vs Phi-4](/fi/compare/aya-23-8b-vs-phi-4/)
- [Aya Expanse 8B vs Phi-4](/fi/compare/aya-expanse-8b-vs-phi-4/)
- [NVIDIA Cosmos Reason2 2B vs Phi 3 Mini 128K Instruct](/fi/compare/cosmos-reason2-2b-vs-phi-3-mini-128k-instruct/)
- [DeepSeek Coder V2 Lite Instruct vs Phi-4](/fi/compare/deepseek-coder-v2-lite-instruct-vs-phi-4/)
- [DeepSeek OCR 2 vs Phi 3 Mini 128K Instruct](/fi/compare/deepseek-ocr-2-vs-phi-3-mini-128k-instruct/)
- [DeepSeek OCR vs Phi 3 Mini 128K Instruct](/fi/compare/deepseek-ocr-vs-phi-3-mini-128k-instruct/)