Roboflow

OpenAI Vision Models

On Roboflow's Vision Evals, GPT-6.2 Sol is OpenAI's top model at 90.3% (#2 of 73). Compare all 20 OpenAI vision models we track, 1 of them open-weight. Try 19 of them on your own images, free in the Roboflow Playground.

20 models · 1 open-weight · 19 free to try · Updated Nov 2026

About OpenAI Vision Models

OpenAI ships vision models on two tiers, and our catalog tracks 20 of them. 1 is open-weight: the weights are downloadable, so you can self-host under their own licenses. The other 19 are proprietary, reached through an API and billed by the provider. The open-weight side runs on MIT terms. The newest additions are GPT-6.2 Astra (Nov 2026) and GPT-6.2 Sol (Nov 2026). The lineup reaches back to CLIP, released Feb 2021.

Between them, the OpenAI models we list cover 14 distinct vision tasks. The widest coverage is classification (20 models), image tagging (20), and captioning (19). For classification, start with GPT-6.2 Astra, GPT-6.2 Sol, or GPT-6.1 Terra.

On the API tier, OpenAI's most recent entry is GPT-6.2 Astra (Nov 2026), billed per token by the provider rather than run on your own hardware. Which tier to start on is a constraints question, not a quality one: reach for the open-weight side when you need offline inference, predictable per-image cost, or a checkpoint you can fine-tune on your own data, and for the API side when you want broad general reasoning without managing GPUs. 19 of the 20 OpenAI vision models run live in the Roboflow Playground, so you can run the same image through several of them and compare the answers before committing to one.

Which OpenAI Model Should You Use?

What each of the 20 OpenAI vision models in our catalog is built for, and how you run it.

GPT-6.2 Astra
Best for: Promptable Concept Segmentation and Instance Segmentation. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6.2 Sol
Best for: Promptable Concept Segmentation and Instance Segmentation. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6.1 Terra
Best for: Promptable Concept Segmentation and Object Detection. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6.1 Sol
Best for: Promptable Concept Segmentation and Object Detection. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6 Luna
Best for: Promptable Concept Segmentation and Object Detection. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6 Sol
Best for: Promptable Concept Segmentation and Object Detection. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-6 Astra
Best for: Promptable Concept Segmentation and Object Detection. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.6 Luna
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.6 Sol
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.6 Terra
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.5
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.4 Mini
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.4 Nano
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.4
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.2
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5.1
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5 Mini
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
GPT-5 Nano
Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through OpenAI's API. Runnable in the Playground.
CLIP
Best for: Image Embedding and Image Similarity. Open weights: MIT license. Self-host it or run it here.

Open-Source OpenAI Models

1 model with downloadable weights you can self-host under their licenses (MIT).

Actions
OpenAI
CLIPOpenAI
—MITFeb 2021

OpenAI Models via API

19 proprietary models where the weights aren't downloadable: access is through each provider's API and billed by them. Try all of them free in the Playground.

Actions
OpenAINEW
$8.00$40.001.1MNov 2026
Try
OpenAINEW
$2.00$10.001.1MNov 2026
Try
OpenAI
$1.00$5.001.1MOct 2026
Try
OpenAI
$2.00$10.001.1MSep 2026
Try
OpenAI
$0.10$0.501.1MSep 2026
Try
OpenAI
GPT-6 SolOpenAI
$2.00$10.001.1MSep 2026
Try
OpenAI
$10.00$50.001.1MSep 2026
Try
OpenAI
$0.20$1.201.5MJul 2026
Try

Frequently Asked Questions About OpenAI Vision Models

Are OpenAI vision models open source?

Partly. 1 of the 20 OpenAI vision models we track is open-weight (CLIP), downloadable and self-hostable under its license (MIT). The other 19 are proprietary and reached through OpenAI's API.

What is the best OpenAI vision model?

On Roboflow's Vision Evals, GPT-6.2 Sol is the highest-scoring OpenAI model, ranked #2 of 73 overall at 90.3%. The next OpenAI model is GPT-6.2 Astra, at #3 (90.2%). The overall score averages the 5 benchmark tasks, so check the task leaderboards if you need one capability.

How many OpenAI vision models are on Roboflow Playground?

We track 20 live OpenAI vision models, 1 open-weight and 19 available through an API. The most recent addition is GPT-6.2 Astra, released Nov 2026.

Can I try OpenAI vision models for free?

Yes. 19 of the 20 OpenAI models run live in the Roboflow Playground. Upload your own image, run several models on it at once, and compare the outputs side by side. You can try it instantly, and a free account gives you more runs with free Roboflow credits.

This page lists all 20 OpenAI vision models in the Roboflow Playground catalog: 1 open-weight model you can self-host and 19 proprietary models accessed through an API. They cover classification, image tagging, and captioning, among other tasks. 19 of them run live in the Roboflow Playground on your own images. Compare licenses, parameters, prices, and release dates side by side, or open any model page for full details.