Title: Gpt Oss Safeguard 20B vs gpt-oss-120b: Parametre, Licens %E2%80%94 LLM EU
Canonical: https://llmeu.com/da/compare/gpt-oss-safeguard-20b-vs-gpt-oss-120b/
Markdown mirror: https://llmeu.com/da/compare/gpt-oss-safeguard-20b-vs-gpt-oss-120b.md
Machine-readable: https://llmeu.com/llms.txt (index) - https://llmeu.com/llms-full.txt (full catalogue) - https://llmeu.com/llms-catalogue.json

# Gpt Oss Safeguard 20B vs gpt-oss-120b

S%C3%A6t to til fire modeller side om side p%C3%A5 de felter, der afg%C3%B8r en udrulning: kontekst, pris, region, licens, sprog, hosting og hvad der er verificeret.

Hvor to modeller adskiller sig, viser denne tabel forskellen og stopper. LLM EU offentligg%C3%B8r ikke et vinderbadge uden en dokumenteret score for den opgave, og et manglende tal forbliver manglende i stedet for at blive et nul.

## Hvorn%C3%A5r man bruger den

### [Gpt Oss Safeguard 20B](/da/models/gpt-oss-safeguard-20b/)

#### Styrker

- 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
- instruction-tuned checkpoint with a chat template in the repository

#### Begr%C3%A6nsninger

- 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 21.5B parameters, self-hosting needs a single accelerator with enough memory for the weights
- a safety classifier built on the gpt-oss architecture: it labels content against a policy supplied at inference time rather than answering questions

#### Opgaver

[F%C3%B8lsomme data](/da/tasks/private/), [Kundesupport](/da/tasks/customer-support/)

### [gpt-oss-120b](/da/models/gpt-oss-120b/)

#### Styrker

- Apache-2.0 licence with no user threshold
- runs on a single 80GB accelerator
- configurable reasoning effort with full chain-of-thought access
- native function calling, browsing and structured output

#### Begr%C3%A6nsninger

- requires the harmony response format; other prompt templates degrade it
- 117B parameters must be resident even though only 5.1B are active
- chain-of-thought output is not intended for end users

#### Opgaver

[R%C3%A6sonnement](/da/tasks/reasoning/), [Kodning](/da/tasks/coding/), [Genfinding (RAG)](/da/tasks/rag/), [Resum%C3%A9](/da/tasks/summarization/)

**Opgaver:** Ikke verificeret: udbyderen har ikke dokumenteret denne p%C3%A5stand.

## Sammenlign modeller

- [Command A+ (05-2026) vs gpt-oss-120b](/da/compare/command-a-plus-05-2026-vs-gpt-oss-120b/)
- [Command R (08-2024) vs Gpt Oss Safeguard 20B](/da/compare/command-r-08-2024-vs-gpt-oss-safeguard-20b/)
- [GLM 4.5 Air vs gpt-oss-120b](/da/compare/glm-4-5-air-vs-gpt-oss-120b/)
- [GLM 4.5V vs gpt-oss-120b](/da/compare/glm-4-5v-vs-gpt-oss-120b/)
- [gpt-oss-20b vs Gpt Oss Safeguard 20B](/da/compare/gpt-oss-20b-vs-gpt-oss-safeguard-20b/)
- [Hunyuan A13B Instruct vs gpt-oss-120b](/da/compare/hunyuan-a13b-instruct-vs-gpt-oss-120b/)