Title: Gpt Oss Safeguard 20B vs gpt-oss-120b: %CE%A0%CE%B1%CF%81%CE%AC%CE%BC%CE%B5%CF%84%CF%81%CE%BF%CE%B9, %CE%86%CE%B4%CE%B5%CE%B9%CE%B1 %E2%80%94 LLM EU Canonical: https://llmeu.com/el/compare/gpt-oss-safeguard-20b-vs-gpt-oss-120b/ Markdown mirror: https://llmeu.com/el/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 %CE%92%CE%AC%CE%BB%CF%84%CE%B5 %CE%B4%CF%8D%CE%BF %CE%AD%CF%89%CF%82 %CF%84%CE%AD%CF%83%CF%83%CE%B5%CF%81%CE%B1 %CE%BC%CE%BF%CE%BD%CF%84%CE%AD%CE%BB%CE%B1 %CE%B4%CE%AF%CF%80%CE%BB%CE%B1-%CE%B4%CE%AF%CF%80%CE%BB%CE%B1 %CF%83%CF%84%CE%B1 %CF%80%CE%B5%CE%B4%CE%AF%CE%B1 %CF%80%CE%BF%CF%85 %CE%BA%CE%B1%CE%B8%CE%BF%CF%81%CE%AF%CE%B6%CE%BF%CF%85%CE%BD %CE%BC%CE%B9%CE%B1 %CE%B1%CE%BD%CE%AC%CF%80%CF%84%CF%85%CE%BE%CE%B7: %CF%83%CF%85%CE%BC%CF%86%CF%81%CE%B1%CE%B6%CF%8C%CE%BC%CE%B5%CE%BD%CE%B1, %CF%84%CE%B9%CE%BC%CE%AE, %CF%80%CE%B5%CF%81%CE%B9%CE%BF%CF%87%CE%AE, %CE%AC%CE%B4%CE%B5%CE%B9%CE%B1, %CE%B3%CE%BB%CF%8E%CF%83%CF%83%CE%B5%CF%82, %CF%86%CE%B9%CE%BB%CE%BF%CE%BE%CE%B5%CE%BD%CE%AF%CE%B1 %CE%BA%CE%B1%CE%B9 %CF%84%CE%B9 %CE%B5%CE%AF%CE%BD%CE%B1%CE%B9 %CE%B5%CF%80%CE%B1%CE%BB%CE%B7%CE%B8%CE%B5%CF%85%CE%BC%CE%AD%CE%BD%CE%BF. %CE%8C%CF%80%CE%BF%CF%85 %CE%B4%CF%8D%CE%BF %CE%BC%CE%BF%CE%BD%CF%84%CE%AD%CE%BB%CE%B1 %CE%B4%CE%B9%CE%B1%CF%86%CE%AD%CF%81%CE%BF%CF%85%CE%BD, %CE%B1%CF%85%CF%84%CF%8C%CF%82 %CE%BF %CF%80%CE%AF%CE%BD%CE%B1%CE%BA%CE%B1%CF%82 %CE%B4%CE%B5%CE%AF%CF%87%CE%BD%CE%B5%CE%B9 %CF%84%CE%B7 %CE%B4%CE%B9%CE%B1%CF%86%CE%BF%CF%81%CE%AC %CE%BA%CE%B1%CE%B9 %CF%83%CF%84%CE%B1%CE%BC%CE%B1%CF%84%CE%AC. %CE%A4%CE%BF LLM EU %CE%B4%CE%B5%CE%BD %CE%B4%CE%B7%CE%BC%CE%BF%CF%83%CE%B9%CE%B5%CF%8D%CE%B5%CE%B9 %CF%83%CE%AE%CE%BC%CE%B1 %CE%BD%CE%B9%CE%BA%CE%B7%CF%84%CE%AE %CF%87%CF%89%CF%81%CE%AF%CF%82 %CF%84%CE%B5%CE%BA%CE%BC%CE%B7%CF%81%CE%B9%CF%89%CE%BC%CE%AD%CE%BD%CE%B7 %CE%B2%CE%B1%CE%B8%CE%BC%CE%BF%CE%BB%CE%BF%CE%B3%CE%AF%CE%B1 %CE%B3%CE%B9%CE%B1 %CE%B1%CF%85%CF%84%CE%AE%CE%BD %CF%84%CE%B7%CE%BD %CE%B5%CF%81%CE%B3%CE%B1%CF%83%CE%AF%CE%B1, %CE%BA%CE%B1%CE%B9 %CE%BC%CE%B9%CE%B1 %CF%84%CE%B9%CE%BC%CE%AE %CF%80%CE%BF%CF%85 %CE%BB%CE%B5%CE%AF%CF%80%CE%B5%CE%B9 %CF%80%CE%B1%CF%81%CE%B1%CE%BC%CE%AD%CE%BD%CE%B5%CE%B9 %CE%B1%CF%80%CE%BF%CF%8D%CF%83%CE%B1 %CE%B1%CE%BD%CF%84%CE%AF %CE%BD%CE%B1 %CE%B3%CE%AF%CE%BD%CE%B5%CE%B9 %CE%BC%CE%B7%CE%B4%CE%AD%CE%BD. ## %CE%A0%CF%8C%CF%84%CE%B5 %CE%BD%CE%B1 %CF%84%CE%BF %CF%87%CF%81%CE%B7%CF%83%CE%B9%CE%BC%CE%BF%CF%80%CE%BF%CE%B9%CE%AE%CF%83%CE%B5%CF%84%CE%B5 ### [Gpt Oss Safeguard 20B](/el/models/gpt-oss-safeguard-20b/) #### %CE%A0%CE%BB%CE%B5%CE%BF%CE%BD%CE%B5%CE%BA%CF%84%CE%AE%CE%BC%CE%B1%CF%84%CE%B1 - 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 #### %CE%A0%CE%B5%CF%81%CE%B9%CE%BF%CF%81%CE%B9%CF%83%CE%BC%CE%BF%CE%AF - 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 #### %CE%95%CF%81%CE%B3%CE%B1%CF%83%CE%AF%CE%B5%CF%82 [%CE%95%CF%85%CE%B1%CE%AF%CF%83%CE%B8%CE%B7%CF%84%CE%B1 %CE%B4%CE%B5%CE%B4%CE%BF%CE%BC%CE%AD%CE%BD%CE%B1](/el/tasks/private/), [%CE%A5%CF%80%CE%BF%CF%83%CF%84%CE%AE%CF%81%CE%B9%CE%BE%CE%B7 %CF%80%CE%B5%CE%BB%CE%B1%CF%84%CF%8E%CE%BD](/el/tasks/customer-support/) ### [gpt-oss-120b](/el/models/gpt-oss-120b/) #### %CE%A0%CE%BB%CE%B5%CE%BF%CE%BD%CE%B5%CE%BA%CF%84%CE%AE%CE%BC%CE%B1%CF%84%CE%B1 - 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 #### %CE%A0%CE%B5%CF%81%CE%B9%CE%BF%CF%81%CE%B9%CF%83%CE%BC%CE%BF%CE%AF - 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 #### %CE%95%CF%81%CE%B3%CE%B1%CF%83%CE%AF%CE%B5%CF%82 [%CE%A3%CF%85%CE%BB%CE%BB%CE%BF%CE%B3%CE%B9%CF%83%CF%84%CE%B9%CE%BA%CE%AE](/el/tasks/reasoning/), [%CE%9A%CF%89%CE%B4%CE%B9%CE%BA%CE%BF%CF%80%CE%BF%CE%AF%CE%B7%CF%83%CE%B7](/el/tasks/coding/), [%CE%91%CE%BD%CE%AC%CE%BA%CF%84%CE%B7%CF%83%CE%B7 (RAG)](/el/tasks/rag/), [%CE%A0%CE%B5%CF%81%CE%AF%CE%BB%CE%B7%CF%88%CE%B7](/el/tasks/summarization/) **%CE%95%CF%81%CE%B3%CE%B1%CF%83%CE%AF%CE%B5%CF%82:** %CE%9C%CE%B7 %CE%B5%CF%80%CE%B1%CE%BB%CE%B7%CE%B8%CE%B5%CF%85%CE%BC%CE%AD%CE%BD%CE%BF: %CE%BF %CF%80%CE%AC%CF%81%CE%BF%CF%87%CE%BF%CF%82 %CE%B4%CE%B5%CE%BD %CE%AD%CF%87%CE%B5%CE%B9 %CF%84%CE%B5%CE%BA%CE%BC%CE%B7%CF%81%CE%B9%CF%8E%CF%83%CE%B5%CE%B9 %CE%B1%CF%85%CF%84%CF%8C%CE%BD %CF%84%CE%BF%CE%BD %CE%B9%CF%83%CF%87%CF%85%CF%81%CE%B9%CF%83%CE%BC%CF%8C. ## %CE%A3%CF%8D%CE%B3%CE%BA%CF%81%CE%B9%CF%83%CE%B7 %CE%BC%CE%BF%CE%BD%CF%84%CE%AD%CE%BB%CF%89%CE%BD - [Command A+ (05-2026) vs gpt-oss-120b](/el/compare/command-a-plus-05-2026-vs-gpt-oss-120b/) - [Command R (08-2024) vs Gpt Oss Safeguard 20B](/el/compare/command-r-08-2024-vs-gpt-oss-safeguard-20b/) - [GLM 4.5 Air vs gpt-oss-120b](/el/compare/glm-4-5-air-vs-gpt-oss-120b/) - [GLM 4.5V vs gpt-oss-120b](/el/compare/glm-4-5v-vs-gpt-oss-120b/) - [gpt-oss-20b vs Gpt Oss Safeguard 20B](/el/compare/gpt-oss-20b-vs-gpt-oss-safeguard-20b/) - [Hunyuan A13B Instruct vs gpt-oss-120b](/el/compare/hunyuan-a13b-instruct-vs-gpt-oss-120b/)