​
Onyx Power Rankings:

OpenAI gpt-4o

OpenAI’s flagship model has been a consisstent FM leveraged for use cases spanning Consumer Lifestyle, Healthtech, and even GovTech applications on Onyx

Claude 3.5 Sonnet

With the release of 3.5, Claude has been able to reclaim the crown of having the best performing FM.

Llama-3 80B

Llama 3 is the king of OSS models in terms of usage. Onyx customers can leverage the power of this model via API or through docker enabled deployment to the Cloud.

​
Open AI GPT

ModelDescriptionContext Window
gpt-4oOpenAI’s most advanced, multimodal flagship model that’s cheaper and faster than GPT-4 Turbo128,000 Tokens
gpt-4o-miniOpenAI’s first SLM that’s cheaper and faster than 3.5-turbo128,000 Tokens
gpt4-turboThe latest GPT-4 Turbo model with vision capabilities.128,000 Tokens
gpt-3.5-turboThe latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and function calling16,385 Tokens

​
Anthropic Claude

ModelDescriptionContext Window
claude-3.5-sonnetOpenAI’s most advanced, multimodal flagship model that’s cheaper and faster than GPT-4 Turbo128,000 Tokens
claude-3-opusThe latest GPT-4 Turbo model with vision capabilities.128,000
claude-3-sonnetThe latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and function calling16,385
claude-3-haikuThe latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and function calling16,385

​
Meta Llama

ModelDescriptionContext Window
Meta-Llama-3-70BMeta’s most advanced OSS model. Deploy via API or host your own via docker deployment to your CSP of choice8,000 Tokens
Meta-Llama-3-8BCapable and fast model that is often fine tuned to optimize inference for hyperspecialized use cases8,000 tokens

​
Google Gemini

ModelDescriptionContext Window
gemini-1.5-flashGoogle’s fastest and most optimized multimodal LLM1,000,000 Tokens
gemini-1.5-proGoogle’s most capable and powerful multimodal LLM boasting the largest context window to date2,000,000 Tokens
If you require any model not listed above, reach out system@abstractionlabs.ai and our team will work on providing access to the model within 30 minutes of first response.