Title: Gemma 4 26B A4B IT vs Gemma 4 31B: Parametrar, Licens %E2%80%94 LLM EU
Canonical: https://llmeu.com/sv/compare/gemma-4-26b-a4b-it-vs-gemma-4-31b/
Markdown mirror: https://llmeu.com/sv/compare/gemma-4-26b-a4b-it-vs-gemma-4-31b.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# Gemma 4 26B A4B IT vs Gemma 4 31B

St%C3%A4ll tv%C3%A5 till fyra modeller sida vid sida p%C3%A5 de f%C3%A4lt som avg%C3%B6r en drifts%C3%A4ttning: kontext, pris, region, licens, spr%C3%A5k, hosting och vad som %C3%A4r verifierat.

D%C3%A4r tv%C3%A5 modeller skiljer sig visar denna tabell skillnaden och stannar. LLM EU publicerar ingen vinnarutm%C3%A4rkelse utan en dokumenterad po%C3%A4ng f%C3%B6r den uppgiften, och en saknad siffra f%C3%B6rblir saknad i st%C3%A4llet f%C3%B6r att bli en nolla.

## N%C3%A4r den ska anv%C3%A4ndas

### [Gemma 4 26B A4B IT](/sv/models/gemma-4-26b-a4b-it/)

#### Styrkor

- mid-size option with the 256K-token window of the larger model
- Apache-2.0 licence
- text and image input
- vendor reports near-flagship scores at a smaller footprint

#### Begr%C3%A4nsningar

- the "A4B" suffix is the vendor%E2%80%99s own designation and is not defined on the model card
- scores are self-reported
- no evaluation has been run by LLM EU

#### Uppgifter

[Resonemang](/sv/tasks/reasoning/), [Kodning](/sv/tasks/coding/), [Vision](/sv/tasks/vision/), [Sammanfattning](/sv/tasks/summarization/), [L%C3%A5g kostnad](/sv/tasks/cheap/)

### [Gemma 4 31B](/sv/models/gemma-4-31b/)

#### Styrkor

- 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

#### Begr%C3%A4nsningar

- 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 32.7B parameters, self-hosting needs a high-memory 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

#### Uppgifter

[Flerspr%C3%A5kigt st%C3%B6d](/sv/tasks/multilingual-support/)

**Uppgifter:** Inte verifierad: leverant%C3%B6ren har inte dokumenterat detta p%C3%A5st%C3%A5ende.

## J%C3%A4mf%C3%B6r modeller

- [Command R (08-2024) vs Gemma 4 26B A4B IT](/sv/compare/command-r-08-2024-vs-gemma-4-26b-a4b-it/)
- [DeepSeek Coder V2 Lite Instruct vs Gemma 4 26B A4B IT](/sv/compare/deepseek-coder-v2-lite-instruct-vs-gemma-4-26b-a4b-it/)
- [DeepSeek R1 Distill Qwen 14B vs Gemma 4 26B A4B IT](/sv/compare/deepseek-r1-distill-qwen-14b-vs-gemma-4-26b-a4b-it/)
- [DeepSeek-R1-Distill-Qwen-32B vs Gemma 4 26B A4B IT](/sv/compare/deepseek-r1-distill-qwen-32b-vs-gemma-4-26b-a4b-it/)
- [DeepSeek V2 Lite vs Gemma 4 26B A4B IT](/sv/compare/deepseek-v2-lite-vs-gemma-4-26b-a4b-it/)
- [Devstral Small 2 (24B) vs Gemma 4 26B A4B IT](/sv/compare/devstral-small-2-24b-vs-gemma-4-26b-a4b-it/)