GPT OSS Safeguard 120B

Tool callingStructured outputReasoning

Website

Safety model for policy screening, moderation, and risk-aware routing workflows

Context: 131K
Released: 11mo ago
Parameters: 120B

Sources:models.dev

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

No benchmark results have been matched to this model yet.

Specifications

Developer: OpenAI
Model family: GPT
Type: Language model
Predecessor: —
Successor: —
Released: October 29, 2025
Model ID: openai/gpt-oss-safeguard-120b
Openness: Open weights
License: —
Parameters: 120B
Input modalities: text
Output modalities: text
Modality: text->text
Max output tokens: 33K
Hugging Face: —
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Additional source evidence

Source model: openai/gpt-oss-safeguard-120b
Source family: gpt-oss
Last updated: 2025-10-29
Model context: 131K
Model output limit: 33K
Open weights: Yes
File attachments: No
Reasoning: Yes
Tool calling: Yes
Structured output: Yes
Temperature control: Yes

Downloadable weights

Hugging Face

Complete source record (JSON)

{
  "id": "openai/gpt-oss-safeguard-120b",
  "name": "GPT OSS Safeguard 120B",
  "description": "Safety model for policy screening, moderation, and risk-aware routing workflows",
  "family": "gpt-oss",
  "last_updated": "2025-10-29",
  "attachment": false,
  "reasoning": true,
  "tool_call": true,
  "structured_output": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 131072,
    "output": 32768
  },
  "release_date": "2025-10-29",
  "open_weights": true,
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/openai/gpt-oss-safeguard-120b"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider endpoint data will appear after this model's first OpenRouter endpoint sync.

Text provenance

EU · product-scoped

OpenAI textGrain watermarking

In the EU, ChatGPT-generated text uses textGrain. OpenAI says API customers can opt in globally for select models, so watermarking depends on the product surface and configuration.

Official explanation

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware