OpenAI vs NLP Cloud

Usage Restrictions

OpenAI is very restrictive about the kind of applications it allows. Until November 2022 you had to explicitly ask OpenAI to review and approve your application before going into production. Since then, here is what they say: "We no longer require you to register your applications with OpenAI. Instead, we'll be using a combination of automated and manual methods to monitor for policy violations."

Some applications are simply not allowed by default, like applications based on paraphrasing and rewriting (considered as "plagiarism"), multi-level marketing, sexual discussions... Moreover, they often require that you explicitly tell your end users that you are using AI in your application (this is the case for summarization, news generation, and anything related to the medical, legal, and financial industries). Here are the OpenAI's usage policies:

At NLP Cloud we believe that our customers should use their best judgement to decide how to use our AI models based on their local laws and their personal ethics. We think it is not our role to inspect and filter our customer's requests.

Data Privacy

There is a significant difference between OpenAI and NLP Cloud when it comes to data privacy.

OpenAI does many things out of customers' data. Data sent to OpenAI is reviewed and used for all sorts of things. The most important part is that your data is stored for an unlimited period and used to train and improve OpenAI's AI models like ChatGPT. See an extract from OpenAI's privacy policy below:

These privacy considerations can be critical for many businesses, especially those dealing with data-sensitive industries like healthcare, legal, finance...

NLP Cloud is HIPAA / GDPR / CCPA compliant, and working on the SOC 2 certification. We cannot see your data, we do not store your data, and we do not use your data to train our own AI models.

Pricing Differences

OpenAI and NLP Cloud both propose pay-as-you-go prices. It means that you can pay after the fact, only for the number of requests or tokens you actually consumed.

At NLP Cloud we also propose standard packages paid upfront. These plans give you access to a specific number of requests per minute. These plans are more cost-effective than pay-as-you-go if you have a large volume of requests to perform as they help us predict your consumption and better scale our service accordingly in advance.

See NLP Cloud's pricing here.

See OpenAI's pricing below:

Let's make a simple simulation. ChatDolphin is equivalent to ChatGPT (GPT 3.5 Turbo), so we are going to compare both prices.

On NLP Cloud, making 20 requests per minute on ChatDolphin, using 2048 tokens each, will cost you $699/month (Enterprise GPU plan).

On OpenAI, it will cost you 0.002/1000 x 2048 x 20 x 60 x 24 x 31 = $3,656/month.

The price difference is very significant, and it is actually even more important when comparing other features like fine-tuning, semantic search, embeddings, speech-to-text, text-to-image...

For example, making 5 speech-to-text requests per minute on Whisper Large, each audio file lasting 200 seconds, will cost you $699/month (Enterprise GPU plan).

On OpenAI it will cost you (0.006/60) x 200 x 5 x 60 x 24 x 31 = $4,464/month.

One last example with image generation. At NLP Cloud we use Stable Diffusion, which is equivalent to OpenAI DALL-E. Making 5 image generation requests per minute on Stable Diffusion in 768 x 768 px format will cost you $699/month (Enterprise GPU plan).

On OpenAI, using the 512 x 512 px format (which is lower than 768 x 768) it will cost you 0.018 x 5 x 60 x 31 x 24 = $4,017/month.

Features And Models Available

OpenAI and NLP Cloud adopted 2 very different strategies: OpenAI builds a couple of in-house models (GPT-3, ChatGPT, GPT-4, DALL-E...) while at NLP Cloud we mix the best open-source AI models (Bart, MPNet, Stable Diffusion...) with our own in-house models (Dolphin, ChatDolphin, ...) on the same platform.

It means that on NLP Cloud you can of course use some ChatGPT competitors like GPT-NeoX and ChatDolphin, but you can also use many other models like Bart, T5, Distilbert, spaCy, etc.

Using specialized smaller models is often much more cost effective and much faster than using huge generative AI models.

Sometimes, some use cases just cannot be covered by regular AI models. That's the case with multilingual AI for example. If you want to use AI in 200 languages, you can use NLP Cloud's multilingual addon (free of charge). OpenAI does not propose so many languages for the moment.

Leveraging specialized models is also a good way to decrease complexity. For example, performing summarization with GPT-3 or ChatGPT will require some prompt engineering, while you can very simply get advanced results thanks to dedicated pre-trained models like Facebook's Bart Large CNN.

Last of all, NLP Cloud proposes an "asynchronous" mode, meaning that you can trigger long running background jobs. It is useful for example if you want to summarize very large documents (up to 1 million tokens), or transcribe very long audio files. OpenAI doesn't propose such a feature.

Reliability, Quota, And Quality Of Support

When comparing OpenAI and NLP Cloud we encourage you to look into 3 additional aspects: reliability of the API, quota, and quality of support.

OpenAI has experienced many severe service interruptions these last months, especially since the release of ChatGPT and GPT-4. They also faced bugs that led to critical security breaches recently, which scared many customers.

Learn more about recent OpenAI security issues here.

At NLP Cloud, we provide a robust infrastructure and we haven't faced any significant service outage since March 2020 (caused by a fire in an OVH datacenter).

Also, it is important to note that OpenAI automatically applies severe quotas to new users. Basically, if you are a new user, you will need to wait for a while and apply for additional quotas before moving your application to production. At NLP Cloud we do not apply such quotas, so you can go to production instantly.

Last of all, it is important to note that it is in general quite difficult to get a prompt and elaborate response from the OpenAI support. At NLP Cloud, our strategy has always been to provide advanced expertise to all our customers. So for any question related to prompt engineering, model selection, fine-tuning, etc. please don't hesitate to reach out to us.

Conclusion

Many customers are looking for alternatives to OpenAI, mainly because of their use case restrictions, data privacy policy, pricing, and service disruptions.

We do believe that the NLP Cloud API is a very good alternative to OpenAI!

At NLP Cloud, we are proud to provide a high-level support to all our customers, and we are constantly adding more cutting-edge AI models in order for our customers to deliver their AI project in no time.

Would you like to have a try? Test NLP Cloud here!

Juliette
Marketing manager at NLP Cloud