Skip to content
ModelCap

gpt-oss-safeguard-20b

Current board #72Open weights

OpenAI·openai/gpt-oss-safeguard-20b

gpt-oss-safeguard-20b is a language model published by OpenAI. 1 OpenRouter endpoint is published in the latest snapshot. It accepts text input and returns text. The published context limit is 131K tokens. The catalogue declares support for reasoning, tool use, structured outputs.

ModelCap Index · Current board
40.6#72
Inherited · 42% support · score interval 40.6–54.8
Market Gravity
18.1
Secondary market signal
Input
$0.075
per 1M tokens
Output
$0.30
per 1M tokens
Context
131K
tokens
Providers
1
endpoint
Gateway spend
—
latest · Vercel share

Providers

Uptime measured over the last 30 minutes
ProviderInputOutputContextQuantUptime
Groq$0.075$0.30131K—100.0%

Hugging Face Inference Providers

Exact repository identity

Provider economics and telemetry published by the Hugging Face Router. These are operational signals, not benchmark results.

ProviderInputOutputContextTTFTThroughputFeatures
featherless-ai
live
——————
groq
live
——131K281 ms704.0 tok/stools · structured

ModelCap Index · Current board

Methodology
40.6
Modeled · lineage · finetune · finetune
Current board
#72
Public rank basis
Modeled · lineage (lineage)
Index support
42.1% · moderate
Index interval
40.6–54.8
Rank posterior
Not available
Top 5 / Top 10 probability
Not available / Not available
Posterior as of
Snapshot timestamp unavailable
Identity binding
Exact catalogue product · openai/openai/gpt-oss-safeguard-20b · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata openai/gpt-oss-safeguard-20b@8a11e17b25c973a24099d4016bf2e17dd7ec1574 (not the evaluation revision)
Direct benchmark observations
0
Breadth-adjusted BES point estimate
44.2
Verified lineage
openai/gpt-oss-20b · minus 1.2 ( finetune)
Bayesian Evidence Score · measured evidence input only

BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.

Observed capability
—
Evidence status
No admitted measured evidence
General preference
—
Coding
—
Agents & tools
—
Reasoning
—
Evidence breadth
0%
Evidence coverage
0%
Market and catalogue signals · secondary, never the language Index rank
Usage (OpenRouter popularity)
13.0
Liquidity (providers × uptime)
16.2
Open reach (HF downloads)
34.6
Surface (context / tools / modalities)
70.2
Freshness
34.1
  • · OpenRouter popularity #142
  • · 1 live provider
  • · HF 30d downloads 97591
  • · Catalogue: context 131072, reasoning, tools
  • · First seen 2025-10-29

These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.

The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.

Experimental capability estimate

Estimate policy
54.7
39.7–69.6 uncertainty interval
Excluded from rank
Confidence
77%
Training support
156 models · 94 lineages
Out-of-domain check
in domain
Feature coverage
100%
Lineage-held-out error
9.1 MAE

Experimental estimate. This layer predicts from admitted evidence and safe metadata only. It never enters ModelCap Score or rank.

Deployment readiness

Metadata index
74.8
partial · 80% metadata coverage
Not capability
Artifact reproducibility
100.0
Access & legal clarity
100.0
Deployability
82.5
Evaluation provenance
0.0

Missing: evaluationProvenance.admitted-evaluation-observations, evaluationProvenance.capability-family-breadth, evaluationProvenance.evaluation-source-diversity, evaluationProvenance.identity-and-source-confidence, evaluationProvenance.evaluation-dataset-revisions.

A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.

Market activity

Venue coverage
Market Gravity
18.1
OpenRouter weekly popularity
#142 of 298

Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.

Arena preference

Awaiting evaluation

This model is visible because it is available in the live catalogue, but it has no trustworthy community result yet. ModelCap does not infer a quality score from price, features, or market activity.

Market Gravity breakdown

Usage55%
13.0

OpenRouter weekly popularity across the full model catalogue.

Liquidity25%
16.2

Independent providers versus the model's open or closed cohort, weighted by uptime.

Open reach15%
34.6

Hugging Face 30-day downloads for open models; neutral for closed models.

Freshness5%
34.1

Time since first public availability, on a six-month half-life.

Specification

Context window
131K tokens
Max output
66K tokens
Inputs
Text
Outputs
Text
Tokenizer
GPT
Cached input tokens
$0.037 / 1M
First seen on OpenRouter
29 Oct 2025

Capabilities

  • Supported: Reasoning
  • Supported: Tool use
  • Supported: Structured output
  • Supported: Response format
  • Not supported: Moderated

Weights & access

Open weights. Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI.

Access
Open weights
Licence
apache-2.0
Pinned revision
8a11e17b25c9
Parameters
21.5B
Architecture
GptOssForCausalLM
Model type
gpt_oss
Hugging Face downloads (30d)
98K
Downloads (all time)
680K
Likes
248
Repository
openai/gpt-oss-safeguard-20b
Repository updated
14 Jan 2026, 18:38 UTC
Base lineage
openai/gpt-oss-20b