Cloudflare Workers AI, the game-changing platform that empowers developers to seamlessly run AI models, is now available in both Free and Paid Workers plans! This article delves into how its pricing scheme works, including the unique "Neuron" concept and a comparison with similar services.
Get Started with Free for Everyone!
Workers AI understands that even budding developers want to experience the power of AI. Therefore, they offer a generous free allocation of 10,000 Neurons per day. With this allowance, you can experiment with various AI models without incurring any costs.
Neuron: The Unit for Measuring AI Usage
Workers AI utilizes "Neuron" as the unit for measuring AI model usage. Each model requires a different number of Neurons to process data. For instance, 10,000 Neurons can be used to:
- Generate 100-200 responses from a large language model (LLM)
- Translate 500 texts
- Transcribe 500 seconds of audio into text
- Perform text classification 10,000 times
- Create 1,500 - 15,000 text embeddings (depending on the model)
With the serverless pricing model, you only pay for what you use. No more hassles of GPU rental, management, or scaling. Curious about how many Neurons your project might need? Try using Workers AI's cost calculator.
Exceed the Free Allowance? No Worries, Paid Plans Await!
The free 10,000 Neuron per day allocation applies to post-beta models. If you require more than that amount, or wish to use beta models, then you'll need to switch to a Paid Workers plan. The cost is $0.011 per 1,000 Neurons used beyond the free allowance.
All Limits Reset Daily
The free allocation and daily usage limits reset every 00:00 UTC. If the limit is exceeded, subsequent operations will fail with an error message.
Beta Models vs. Non-Beta Models
Starting April 1, 2024, certain models will incur charges after the free 10,000 Neuron per day allocation is exceeded. These models are:
- bge-small-en-v1.5
- bge-base-en-v1.5
- bge-large-en-v1.5
- distilbert-sst-2-int8
- llama-2-7b-chat-int8
- llama-2-7b-chat-fp16
- mistral-7b-instruct-v0.1
- m2m100-1.2b
- resnet-50
- whisper
Cloudflare will continue to calculate Neuron costs for other models and bring them out of beta in the future.
Workers AI vs. Competitors: Cost Comparison
Workers AI employs Neurons to measure and charge for AI usage. This might differ from the input-based pricing schemes offered by other providers. To facilitate comparison, Workers AI provides a table of estimated Neuron costs and usage for various available models.
Please note that the following information is for comparison purposes only. All conversions are based on Cloudflare's public pricing as of March 1, 2024, and do not include taxes or other fees.
Example Usage:
Suppose you use 50,000 Neurons per day for a month. The cost would be:
(50,000 Neurons - 10,000 free Neurons) * 30 days * $0.011 per 1,000 Neurons = $13.20
With a generous free allowance and a per-use payment model, Workers AI presents a cost-effective and predictable AI solution. Intrigued to give Workers AI a try? Visit Cloudflare's website to get started!
Bonus: Price Change Archive
Cloudflare is committed to continuously optimizing their platform and delivering value to customers. For instance, on April 2, 2024, they reduced the pricing for mistral-7b-instruct by 17x and llama-2-7b-chat-int8 by 7x. You can check out the "Price Change Archive" table to see how the prices of these models have evolved
0 comments:
Post a Comment