Skip to main content

OpenAIAdapter

Adapter for OpenAI models (GPT-4o, GPT-4, etc.).

Supported Models​

GPT-4.5 Series​

  • gpt-4.5 - Advanced reasoning model with emotional intelligence and improved factual accuracy.
  • gpt-4.5-mini - A smaller, more cost-efficient version of GPT-4.5.

GPT-4.1 Series​

  • gpt-4.1 - Flagship model with 1-million-token context window, strong coding performance.
  • gpt-4.1-mini - Smaller variant of GPT-4.1.
  • gpt-4.1-nano - Ultra-small, highly optimized variant.

o1 Series (Reasoning Models)​

  • o1 - Advanced reasoning model (December 2024)
  • o1-mini - Faster reasoning model
  • o1-preview - Preview of O1 capabilities
  • gpt-4o - Most capable multimodal model (May 2024), real-time text, audio, and image processing.
  • gpt-4o-2024-11-20 - Latest GPT-4o snapshot
  • gpt-4o-2024-08-06 - August 2024 snapshot
  • gpt-4o-2024-05-13 - Initial release snapshot
  • gpt-4o-mini - Fast, affordable multimodal model with 128K context window.
  • gpt-4o-mini-2024-07-18 - Initial mini release

GPT-4 Turbo Series​

  • gpt-4-turbo - Latest GPT-4 Turbo with vision
  • gpt-4-turbo-2024-04-09 - April 2024 snapshot
  • gpt-4-turbo-preview - Preview version
  • gpt-4-0125-preview - January 2024 preview

GPT-4 Series​

  • gpt-4 - Base GPT-4 model (March 2023)
  • gpt-4-0613 - June 2023 snapshot

GPT-3.5 Series​

  • gpt-3.5-turbo - Cost-effective model
  • gpt-3.5-turbo-0125 - January 2024 snapshot

Note: Model availability and naming may change. Refer to OpenAI's model documentation for the latest information.

Usage​

from benchwise import evaluate

@evaluate("gpt-3.5-turbo", "gemini-2.5-flash", temperature=0.7)
async def test_openai(model, dataset):
    responses = await model.generate(dataset.prompts)
    return {"responses": responses}

Authentication​

Set environment variable:

export OPENAI_API_KEY="your-api-key"

See Also​