OpenAIReleased Apr 2025Proprietary
GPT-4.1 is deprecated and can't be run here; try GPT-6.1 Sol (81.7% on Vision Evals) instead.
GPT-4.1 has no score on the current Vision Evals. On the earlier Vision Evals board (last updated July 9, 2026) it passed 70.2% of visual understanding tasks (#19 of 77). It is a proprietary model from OpenAI, so commercial use follows OpenAI's terms.
This model is deprecated
GPT-4.1 and can no longer be run here. Its evaluation results and details remain available for reference. Try GPT-6.1 Sol instead.
GPT-4.1 has been deprecated by its provider and can no longer be evaluated on the current benchmark. The legacy Vision Evals results below are preserved for reference. See the current Vision Evals
| Category | Passed | Score |
|---|---|---|
| Spatial Understanding | 16 / 19 | 84.2% |
| Defect Detection | 12 / 15 | 80% |
| Object Understanding | 11 / 14 | 78.6% |
| Document Understanding | 6 / 9 | 66.7% |
| Object Counting | 2 / 10 | 20% |
| Category | Passed | Score |
|---|---|---|
| License Plate Recognition | 28 / 30 | 93.3% |
| Text Recognition | 26 / 30 | 86.7% |
| Focused Scene OCR | 80 / 99 | 80.8% |
| VQA & Extraction | 47 / 60 | 78.3% |
| Handwritten Math | 5 / 10 | 50% |
Scores based on a single evaluation run · Methodology
View all legacy Vision Evals results →GPT-4.1, released by OpenAI in April 2025, is a multimodal large language model that advances the GPT-4 series with major improvements in coding, reasoning, and instruction following. It accepts both text and images, supports tool calling and structured outputs, and features an expanded context window of up to ~1 million tokens—enabling it to process very large documents, multi-file codebases, or long conversations in a single prompt. Its knowledge is current through June 2024.
GPT-4.1 costs $2.00 per 1M input tokens and $8.00 per 1M output tokens.
Pricing updated Nov 4, 2026
GPT-4.1 is proprietary: the weights are not distributed, and the GPT-4.1 license is the vendor's commercial terms of service that you accept when you call the API.
Vendor terms govern data retention, whether your inputs can be trained on, rate limits, and regional availability, and they can change with notice. Review them if you handle regulated or customer data.
Proprietary terms are set by the vendor rather than negotiated per project, and no open-source obligation attaches to your code. If you would rather deploy an open model instead of GPT-4.1, Roboflow's licensing page lists the models whose commercial license is included when you deploy on Roboflow.
Do not hesitate to reach out with questions for your commercial project. Our team will help you start solving business problems on the first call. See Roboflow commercial licensing for the models included in each plan.
Talk to salesLicense information is provided as a guide and is not legal advice.
Estimated cost per task vs. Visual Understanding score, for this model and others ranked near it. Upper-left is the sweet spot (high quality, low cost). Based on Vision Evals (legacy) results.
10 of 11 models plotted · 1 not yet evaluated
| Model | Score | Median tokens | Est. cost / task | Compare |
|---|---|---|---|---|
| GPT-5.6 Sol | 76.1% | 1.5K | $0.0058 | Compare |
| Gemini 3.1 Pro | 75.8% | 1.1K | $0.0024 | Compare |
| Gemini 3 Flash | 74.6% | 1.4K | $0.0014 | Compare |
| GPT-5 Mini | 73.1% | 1.8K | $0.0006 | Compare |
| Qwen3.5-27B | 71.6% | 1.2K | $0.0003 | Compare |
| GPT-4.1(this model) | 70.2% | 977 | — | — |
| Claude Sonnet 5 | 70.2% | 2.2K | $0.0048 | Compare |
| Claude Sonnet 4.6 | 70.2% | 2.3K | $0.0080 | Compare |
| GPT-5.6 Luna | 70.2% | 1.5K | $0.0003 | Compare |
| Gemini 2.5 Pro | 70.2% | 856 | $0.0060 | Compare |
| Gemini 3.1 Flash-Lite | 68.7% | 1.1K | $0.0003 | Compare |
Other models worth comparing for similar use cases.
Other versions in the same family as GPT-4.1.
Yes. GPT-4.1 accepts image input, and on Roboflow's previous vision benchmark it passed 70.2% of visual understanding tasks (#19 of 77) and scored 81.2% on OCR.
No. GPT-4.1 has been deprecated by its provider and can no longer be run, so it is not part of Roboflow's current Vision Evals. Its results from the previous benchmark are preserved on this page for reference.