DEVELOPER REFERENCE · VERIFIED MODEL EVIDENCE

OpenAI
Compare 12 source-verified model records, repository licenses, published weight formats and dated evidence. No overall model ranking.
Evidence statusSource-first verified
Latest record check2026-11-13
Latest record full verification2026-10-09
DIRECT ANSWER · SOURCE-OBSERVED REGISTRY
What OpenAI model evidence is available?
Open Model Weights currently tracks 12 field-verified OpenAI model records. In this indexed sample, the recorded modalities are 5 Audio → text; 4 Zero Shot Image Classification; 3 Text. The published repository license labels include 7 apache-2.0; 4 Unknown / not declared; 1 mit. The listed weight formats include PyTorch (10), Safetensors (9), ONNX (1).
Coverage describes the model repositories indexed here, not every model released by the publisher. A checked source, an unknown field and an unchanged source revision are different states. No missing value is treated as a negative answer.
12verified model records
3recorded modalities
3distinct license labels
12model repository links
DEVELOPER HQ / PRIMARY ORGANIZATION LOCATION
🇺🇸
Primary organization locationSan Francisco, California, United States
OpenAI
This location describes the developer or primary organization. It is not the nationality of a model, dataset, contributor, training run or infrastructure.
OpenAI publishes a subset of its AI research as downloadable open-weight models, including gpt-oss reasoning checkpoints and older Whisper speech and CLIP vision models. These weights are separate from OpenAI's proprietary hosted GPT service offerings.
This developer profile covers verified downloadable repositories, not OpenAI's complete API catalog. Even when the same publisher offers both hosted and self-hosted products, model identifiers, costs, weights and licenses must be checked separately.
01gpt-oss: downloadable reasoning weights
OpenAI's official open-models page describes gpt-oss-120b and gpt-oss-20b as open-weight reasoning models intended for local and server deployment. Their available model files, MoE activation, runtime support and output controls must be assessed against a specific hardware setup rather than a single model ranking.
02Whisper for speech recognition
Whisper checkpoints are designed for speech-to-text workloads, including multilingual transcription. The register includes different parameter sizes and a large-v3-turbo variant. Check the individual model card and language/audio setup; a model identifier does not by itself specify accuracy for any target dataset.
03CLIP and other non-chat research models
CLIP combines image and text representations for similarity and zero-shot image tasks. The indexed CLIP entries cannot be treated as GPT-style chat systems. Several model-card license labels are still unresolved here; that is missing evidence, not a conclusion that use is prohibited or allowed.
04Open weights are not equivalent to the OpenAI API
OpenAI's official guidance identifies gpt-oss as downloadable Apache-2.0-weight models and explains that they are not served through OpenAI's own API. Running weights yourself creates separate infrastructure cost, model-safety and governance responsibilities; proprietary GPT endpoints have distinct service terms.
Publisher background is attributed to official sources. Model counts and license classifications come from OMW repository evidence; no model benchmark or processing location is inferred.
Textapache-2.0
Larger open-weight reasoning model
Inspect the 120B checkpoint for its MoE artifact, available inference runtimes and Apache-2.0 license.
Textapache-2.0
Smaller open-weight reasoning model
Check the 20B option for lower-scale deployments; do not infer memory from the name alone.
Audio → textmit
Speech transcription
Use Whisper large-v3-turbo for speech-to-text tasks and evaluate accuracy on local language and audio data.
Zero Shot Image ClassificationUnknown / not declared
Image-text feature similarity
Use CLIP as a feature/zero-shot image model and investigate any unresolved repository license evidence.
Textapache-2.0
Task-specific classification
Review the exact privacy-filter model card, license and supported classifications before operational use.
Compare models →Explore weight-memory estimates →Understand licensing →
FIELD-BY-FIELD COMPARISON · PRIMARY LINKS
Model evidence at a glance.
Compare the exact entries in our registry. Parameter values are taken from the source-checked record; a license label is not a legal opinion, and recognized weight files are not an inference-speed benchmark. Open each record for the underlying citations and observed history.
Compare two to four models →Explore weight memory →Browse observed history →
LICENSE & COVERAGE · NOT LEGAL ADVICE
How do the recorded licenses differ?
Different models from the same developer can use different license terms. The distribution below describes the labels and field-level classifications observed in the listed repositories; it does not automatically establish permissions for a specific deployment.
License labels
- apache-2.07 of 12
- Unknown / not declared4 of 12
- mit1 of 12
Commercial-use classifications
- allowed8 of 12
- unknown4 of 12
Model modalities
- Audio → text5 of 12
- Zero Shot Image Classification4 of 12
- Text3 of 12
EVIDENCE METHOD · TRANSPARENT LIMITS
How this developer profile is assembled.
We aggregate the current source-checked model records, rather than publish an unverified corporate biography. Repository metadata, model cards, license files and recognized weight artifacts supply the facts where available. The record-check date does not assert that a model or its license changed on that date.
Source-derived runtime signals are not hardware benchmarks. Weight-only memory estimates omit runtime overhead, activations and other deployment requirements. Model-specific uncertainties remain visible in the individual evidence records.
Evidence policy →Full methodology →Independent measurements →
QUESTIONS · DIRECT ANSWERS
Frequently asked questions about OpenAI models.
Which OpenAI models are covered?
Open Model Weights currently lists 12 field-verified records for OpenAI. Their recorded model modalities are: 5 Audio → text; 4 Zero Shot Image Classification; 3 Text. This is a registry sample, not an exhaustive publisher catalog.
Which licenses are recorded for OpenAI models?
In the 12 indexed records, repository license labels are: 7 apache-2.0; 4 Unknown / not declared; 1 mit. License labels do not replace the model-specific terms or establish rights for every use case.
Where can I verify OpenAI model information?
Each indexed model links to its official source repository and a field-level Open Model Weights profile. The most recent record checks shown here reach 2026-11-13; checking a repository does not imply that its contents changed.
Does OpenAI release any downloadable model weights?
Yes. OpenAI's gpt-oss 120B and 20B checkpoints are presented as downloadable open-weight reasoning models; its Hugging Face organization also hosts Whisper and CLIP research checkpoints.
Can I call gpt-oss through the OpenAI API?
OpenAI says the gpt-oss models are not offered through its own hosted API. They can be run with compatible self-hosted inference software or through third parties under separate terms.
Are all OpenAI model weights Apache-2.0?
No blanket license should be inferred. OpenAI's gpt-oss releases are Apache 2.0, but the checked CLIP registry entries have unresolved license labels and other OpenAI families may differ. Verify the exact checkpoint.
COMPLETE INDEXED SELECTION
All 12 indexed model profiles.