GooseAi
Revolutionize NLP access: cost-effective, fast, easy integration, diverse models.. [Paid]
Last verified:
What is GooseAi?
GooseAI is a fully managed NLP-as-a-Service delivered via API that provides inference for open source language models at 30% the cost of competitors. It makes deploying NLP services easier and more accessible for creative technologists building products on top of large language models. The platform offers a plug-and-play solution for serving open source language models with feature parity to industry standard APIs, requiring only 2 lines of code changes to migrate.
The service provides a Completions API that is a predictive, rule-based text generation tool using prompts to create intelligent predictions and complete text. It supports classic NLP use cases including text completion/generation, with question/answer and classification coming soon. GooseAI runs on a credit system where users pre-purchase credits through a self-serve platform to pay for tokens generated in each API call.
GooseAI is designed for developers, creative technologists, and companies building AI products who want to reduce infrastructure costs. The platform offers an AI Playground environment for casual experimentation with models, where users can try different tunable parameters and evaluate model coherence for their use cases. It is a joint venture by CoreWeave and Antran.
Available models include GPT-Neo variants (125M, 1.3B, 2.7B), GPT-J 6B, GPT-NeoX 20B, and Fairseq models (1.3B, 2.7B, 6.7B, 13B). All models are EleutherAI or Facebook Mixture of Experts models trained on The Pile, RoBERTa, and CC100 subset data with gpt2 or pile tokenizers.
GooseAi pricing
Pricing model: Freemium
GooseAI runs on a credit system where users pre-purchase credits through the self-serve platform. Pricing is based on tokens generated per API call with a base price including first 25 tokens plus additional token fees: GPT-Neo 125M ($0.000035/request + $0.000001/token), GPT-Neo 1.3B ($0.000110/request + $0.000003/token), GPT-Neo 2.7B ($0.000300/request + $0.000008/token), GPT-J 6B ($0.000450/request + $0.000012/token), Fairseq 13B ($0.001250/request + $0.000036/token), GPT-NeoX 20B ($0.002650/request + $0.000063/token). Account creation is free with AI Playground access. Bulk credit purchases at discount available for Enterprise customers.
GooseAi pros
- 30% lower cost than competitors
- 70% cost savings with transparent pricing
- Feature parity with OpenAI API
- Switch by changing just one line of code
- Free account setup in under 1 minute
- Pay only for tokens generated, not input
- Pre-purchase credit system for cost control
- AI Playground for model experimentation
- Multiple open source model sizes available
- Industry's fastest generation speeds
- Up to 2,048 tokens per API call
- Self-serve platform for easy management
- Bulk credit discounts for Enterprise
- Comprehensive API documentation
- Works with standard OpenAI API Client
GooseAi cons
- Only charges output tokens but input not free
- Question/Answer feature coming soon
- Classification feature coming soon
- Limited to open source models only
- No free tier for API usage
- Requires pre-purchasing credits
- Maximum 2,048 tokens per request
- Smaller models may have lower quality
Frequently asked questions about GooseAi
What is GooseAI?
GooseAI is a fully managed inference service delivered via API that makes deploying NLP services easier and more accessible for creative technologists building products on top of large language models. It provides a plug-and-play solution for serving open source language models at the industry's best economics by simply changing 2 lines of code, with feature parity to other well known APIs.
I use OpenAI, how hard is it to switch?
Switching is as easy as changing one line of code. With feature parity to industry standard APIs, your product works the same but faster. You simply add openai.api_base = 'https://api.goose.ai/v1' to your existing OpenAI code to migrate.
What models do you have available?
GooseAI offers multiple open source models including GPT-Neo (125M, 1.3B, 2.7B), GPT-J 6B, GPT-NeoX 20B, and Fairseq models (1.3B, 2.7B, 6.7B, 13B). All are EleutherAI or Facebook Mixture of Experts models trained on The Pile, RoBERTa, and CC100 subset data.
How does cost and performance compare to others?
GooseAI delivers NLP-as-a-Service at 30% the cost of competitors with 70% cost savings. The infrastructure provides industry's fastest generation speeds. Pricing is solely based on tokens generated (not input tokens), giving users more control and transparency compared to competitors who charge for both input and output.
Can I try GooseAI?
Yes, account creation is free and takes less than 1 minute. GooseAI also provides an AI Playground environment for casual experimentation with models on the platform where you can try different tunable parameters and evaluate model coherence for your use case before committing to API usage.
How does your pricing work?
Pricing is based on tokens generated per API call. The base price includes the first 25 tokens generated, and you pay per additional token beyond that. You can configure the maximum number of tokens per API call up to 2,048 tokens to control costs. This output-only pricing gives you control and predictability.
How do I pay to use your API?
GooseAI runs on a credit system where you pre-purchase credits through the self-serve platform to pay for tokens generated in each API call. After creating an account, you generate a secret key from the API Key management page on your Dashboard to access the API. Bulk credit purchases at discount are available for Enterprise customers.
How does your pricing compare to competitors?
Other APIs list pricing for tokens and charge for both input tokens and output tokens, causing compounding costs as previous generations become context. GooseAI prices solely based on tokens generated, giving you 70% cost savings with far more transparency and control over maximum tokens per call.
What's with all the goose puns?
GooseAI embraces the goose theme with puns like 'Meet our gaggle' for models, 'Geese migrate. So should you' for switching, 'Your wild goose chase ends here', and 'take flight' for signing up. The branding plays on the goose name while communicating the migration message from expensive AI infrastructure.
What is the API endpoint format?
The Completions API endpoint is https://api.goose.ai/v1/engines/[model]/completions. You access it using the OpenAI API Client by setting openai.api_base = 'https://api.goose.ai/v1' and your API key. You can list engines, get individual engine info, and create completions using standard OpenAI client methods.