Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

GPT OSS Safeguard 20B

OpenAI · 22B · open weights

Strongest in Secure code (#38 of 274), weakest in Jailbreak resistance (#206 of 272). Above par in 5 of 6 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5.5 and MiniMax M2.7 and ahead of Phi 3 Small 128k Instruct and GPT-5.6 Astra.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Safety52.2#127/336
Secure code62.8#38/274
Fairness52.2#94/299
Toxicity avoidance54.6#124/272
Harm refusal50.4#149/299
Jailbreak resistance44.9#206/272

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for GPT OSS Safeguard 20B, left for the other.

OLMoE 1B-7B Instructwon 6 · lost 0
OLMo 7B Instruct Hfwon 6 · lost 0
RakutenAI 7B Chatwon 6 · lost 0
SmolLM 1.7B Instructwon 6 · lost 0
Qwen2.5 7B Instructwon 6 · lost 0
Gemma 7B Itwon 6 · lost 0
Grok 4.1 Fastwon 6 · lost 0
Laguna Xs 2.1won 6 · lost 0
RakutenAI 7B Instructwon 6 · lost 0
GLM-4 9B Chatwon 6 · lost 0
Qwen2 7B Instructwon 6 · lost 0
Grok 4won 6 · lost 0
Grok 3won 6 · lost 0
Grok Build 0.1won 6 · lost 0
Qwen1.5 14B Chatwon 6 · lost 0
Mimo V2.6 Flashwon 6 · lost 0
Gemini 3.6 Flashwon 6 · lost 0
DeepSeek V4 Flashwon 6 · lost 0
GLM-4.6won 6 · lost 0
Grok 4.3won 6 · lost 0
DeepSeek V3won 6 · lost 0
DeepSeek V3p1won 6 · lost 0
DeepSeek V3.2won 6 · lost 0
Fuguwon 6 · lost 0
Gemma 4 31Bwon 6 · lost 0
DBRX Instructwon 5 · lost 1
PowerMoE 3Bwon 5 · lost 1
Phi 2won 5 · lost 1
H2o Danube3 500M Chatwon 5 · lost 1
PowerLM 3Bwon 5 · lost 1
Smollm2 1.7B Instructwon 5 · lost 1
LFM2 350Mwon 5 · lost 1
Qwen3 30B A3B Instructwon 5 · lost 1
SmolLM2 360M Instructwon 5 · lost 1
Qwen2.5 0.5B Instructwon 5 · lost 1
Jamba Mini 1.7won 5 · lost 1
Magistral Smallwon 5 · lost 1
Aya 23 8Bwon 5 · lost 1
Laguna S 2.1won 5 · lost 1
GPT-3.5 Turbowon 5 · lost 1
Llama 3.1 8B Instructwon 5 · lost 1
H2o Danube3 4B Chatwon 5 · lost 1
Zephyr 7B Betawon 5 · lost 1
Ling 3.0 Flash VLwon 5 · lost 1
Magistral Mediumwon 5 · lost 1
Jamba 1.6 Miniwon 5 · lost 1
Command R7Bwon 5 · lost 1
LFM2 700Mwon 5 · lost 1
Command R Pluswon 5 · lost 1
Command Awon 5 · lost 1
Jamba Large 1.7won 5 · lost 1
NexusRaven V2 13Bwon 5 · lost 1
Ling 3.0 Flashwon 5 · lost 1
Qwen3 8Bwon 5 · lost 1
Open Mixtral 8x7bwon 5 · lost 1
Jamba 1.5 Largewon 5 · lost 1
GLM-5.2won 5 · lost 1
DeepSeek LLM 67B Chatwon 5 · lost 1
Gemini 3.5 Flashwon 5 · lost 1
Phi-4 Mini Instructwon 5 · lost 1
Jamba 1.5 Miniwon 5 · lost 1
Qwen2.5 3B Instructwon 5 · lost 1
Aya 23 35Bwon 5 · lost 1
MiniMax M3won 5 · lost 1
Gemma 4 E2Bwon 5 · lost 1
Jamba 1.6 Largewon 5 · lost 1
GLM-4.5won 5 · lost 1
Kimi K2.6won 5 · lost 1
Jamba Instructwon 5 · lost 1
Qwen3.7 Maxwon 5 · lost 1
SeaLLMs V3 7B Chatwon 5 · lost 1
Qwen2.5 7B Instruct 1Mwon 5 · lost 1
Mistral Smallwon 5 · lost 1
Kimi K2.5won 5 · lost 1
DeepSeek Reasonerwon 5 · lost 1
Qwen2.5 1.5B Instructwon 5 · lost 1
Gemma 4 12Bwon 5 · lost 1
Hunyuan A13B Instructwon 5 · lost 1
Gemma 3 4B Itwon 5 · lost 1
Gemini 2.5 Prowon 5 · lost 1
Gemma 4 26B A4Bwon 5 · lost 1
Gemini 2.5 Flashwon 5 · lost 1
Claude Opus 4.6won 3 · lost 1
Claude Sonnet 4.6won 3 · lost 1
OLMo 2 7B Instructwon 4 · lost 2
Gemma 2B Itwon 4 · lost 2
DeepSeek R1won 4 · lost 2
SmolLM 360M Instructwon 4 · lost 2
Gemma 2 9B Itwon 4 · lost 2
Open Codestral Mambawon 4 · lost 2
LFM2 1.2Bwon 4 · lost 2
QwQ 32Bwon 4 · lost 2
Yi 1.5 9B Chatwon 4 · lost 2
Gemma 4 E4Bwon 4 · lost 2
Sarvam 1won 4 · lost 2
WizardLM 2.8x22Bwon 4 · lost 2
Llama 4 Scout Instructwon 4 · lost 2
Llama 3.2 3B Instructwon 4 · lost 2
Kimi K2.7 Codewon 4 · lost 2
Mimo V2.5 Prowon 4 · lost 2
Gemini 2.0 Flash Litewon 4 · lost 2
Starling Lm 7B Betawon 4 · lost 2
Open Mixtral 8x22bwon 4 · lost 2
Gemini 1.5 Prowon 4 · lost 2
GLM-5.3 Flashwon 4 · lost 2
Apertus 8B Instructwon 4 · lost 2
DeepSeek Chatwon 4 · lost 2
DeepSeek V4 Prowon 4 · lost 2
Qwen2.5 32B Instructwon 4 · lost 2
LongWriter Glm4 9Bwon 4 · lost 2
Gemini 1.5 Flashwon 4 · lost 2
Qwen2 VL 72Bwon 4 · lost 2
GLM-5 Turbowon 4 · lost 2
Llama 3 8B Instructwon 4 · lost 2
Sarvam 30Bwon 4 · lost 2
Gemini 2.0 Flashwon 4 · lost 2
Gemma 3 27B Itwon 4 · lost 2
GPT-5.6 Astrawon 4 · lost 2
SEA LION V1 7B ITwon 3 · lost 3
Komodo 7B Basewon 3 · lost 3
Gemma 2 2B Itwon 3 · lost 3
IBM Granite 4.0 Microwon 3 · lost 3
GPT OSS 20Bwon 3 · lost 3
Llama 3.2 1B Instructwon 3 · lost 3
Claude 3 Haikuwon 3 · lost 3
Qwen2.5 14B Instructwon 3 · lost 3
Llama 3 70B Chat HFwon 3 · lost 3
Kimi K2won 3 · lost 3
Mimo V2.6 Prowon 3 · lost 3
Kimi K2 Thinkingwon 3 · lost 3
Gemma 3 12B Itwon 3 · lost 3
SeaLLM 7B V2won 3 · lost 3
GPT-6 Solwon 3 · lost 3
Mistral Largewon 3 · lost 3
K2 Chatwon 3 · lost 3
GPT-4.1 Nanowon 3 · lost 3
Phi 3 Mini 4k Instructwon 3 · lost 3
Nova Prowon 3 · lost 3
Krutrim 2 Instructwon 3 · lost 3
GLM-5.1won 3 · lost 3
EAI SageAlign 8B LoRAwon 3 · lost 3
GPT-4.1 Miniwon 3 · lost 3
GPT-4won 3 · lost 3
O4 Miniwon 3 · lost 3
GPT-5.6 Lunawon 3 · lost 3
Nova Litewon 3 · lost 3
Internlm2 Chat 20Bwon 3 · lost 3
Fugu Ultra V2won 3 · lost 3
O3 Miniwon 3 · lost 3
GPT-4.1won 3 · lost 3
Gemini 3.8 Flashwon 2 · lost 4
Apertus 70B Instructwon 2 · lost 4
Granite 4.0 H Microwon 2 · lost 4
Granite 4.0 H Tinywon 2 · lost 4
Llama 3 70B Instructwon 2 · lost 4
Phi 3.5 Mini Instructwon 2 · lost 4
GPT-4o Miniwon 2 · lost 4
Qwen3.8 Maxwon 2 · lost 4
Llama 3.1 Tulu 3 8Bwon 2 · lost 4
Qwen Maxwon 2 · lost 4
GPT-6 Lunawon 2 · lost 4
Phi 3.5 MoE Instructwon 2 · lost 4
Nova Microwon 2 · lost 4
Llama 3 8B Chat Hfwon 2 · lost 4
C4ai Aya Expanse 32Bwon 2 · lost 4
C4ai Aya Expanse 8Bwon 2 · lost 4
Llama 3.2 Vision (11B)won 2 · lost 4
GPT OSS 120Bwon 2 · lost 4
Titan Text Express V1won 2 · lost 4
Qwen2.5 72B Instructwon 2 · lost 4
Llama 2 70B Chat Hfwon 2 · lost 4
Llama 2 13B Chat Hfwon 2 · lost 4
Llama 3 8B Instruct RRwon 2 · lost 4
Granite 4.0 H Smallwon 2 · lost 4
Llama 2 7B Chat Hfwon 2 · lost 4
Claude 3.7 Sonnetwon 2 · lost 4
Qwen2 72B Instructwon 2 · lost 4
Claude 3 Sonnetwon 2 · lost 4
Claude Opus 4.7won 2 · lost 4
Meta SecAlign 8Bwon 2 · lost 4
Phi 4won 2 · lost 4
Muse Spark 1.1won 2 · lost 4
GPT-5.6 Solwon 1 · lost 5
Qwen3.5 35B A3Bwon 1 · lost 5
Qwen3.5 27Bwon 1 · lost 5
Qwen3.5 122B A10Bwon 1 · lost 5
Nova Premierwon 1 · lost 5
Grok 4.5won 1 · lost 5
GPT-5.5won 1 · lost 5
Sarvam 105Bwon 1 · lost 5
GPT-4owon 1 · lost 5
GPT-5.4won 1 · lost 5
Claude Haiku 4.5won 1 · lost 5
GPT-5.6 Terrawon 1 · lost 5
Cogito 671B V2 P1won 1 · lost 5
Claude Opus 4won 1 · lost 5
Claude Sonnet 4.5won 1 · lost 5
O1 Miniwon 1 · lost 5
Claude Sonnet 4won 1 · lost 5
GPT-4 Turbowon 1 · lost 5
GPT-5.1won 1 · lost 5
O3won 1 · lost 5
GPT-5.2won 1 · lost 5
O1won 1 · lost 5
Grok 4.6won 1 · lost 5
GPT-5 Miniwon 1 · lost 5
MiniMax M2won 1 · lost 5
Claude 3.5 Haikuwon 1 · lost 5
Claude 3 Opuswon 1 · lost 5
Claude Opus 4.8won 1 · lost 5
Claude 3.5 Sonnetwon 1 · lost 5
Hy3won 1 · lost 5
Claude Sonnet 5won 1 · lost 5
Gemma 2 27B Itwon 0 · lost 6
Qwen3.5 397B A17Bwon 0 · lost 6
Apriel 1.5 15B Thinkerwon 0 · lost 6
Claude Opus 5won 0 · lost 6
GPT-5won 0 · lost 6
GPT-5 Nanowon 0 · lost 6
Inklingwon 0 · lost 6
Grok 4.7won 0 · lost 6
MiniMax M2.7won 0 · lost 6
Claude Opus 5.5won 0 · lost 6

§ 3 · Sources

Where the numbers come from

1 publication, 6 figures. Every one links to the page it was read from.

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-27
Enkrypt · Jailbreak risk 17.1%Enkrypt · Harmful content risk 25.6%Enkrypt · CBRN risk 18%Enkrypt · Toxicity risk 3.1%Enkrypt · Bias risk 74.4%Enkrypt · Insecure code risk 7.6%

Badge

PublicAI Index badge for GPT OSS Safeguard 20B[![PublicAI Index](https://publicai.io/model-index/badge?model=gpt-oss-safeguard-20b)](https://publicai.io/model-index/m/gpt-oss-safeguard-20b)