OpenAI
AI models for text generation, embeddings, image generation, speech-to-text, and more
Tags: ml, ai, ai-models, nlp
Category: Machine Learning
Use Cases
- Primary LLM provider for OpenClaw chat completions and tool use
- Generate images with DALL-E 3 via OpenClaw's image generation skill
- Transcribe audio with Whisper and generate speech with TTS models
Tips
- Use GPT-4o-mini as default model and escalate to GPT-4o only when reasoning quality matters
- Set up usage limits in the OpenAI dashboard to prevent unexpected bills
- Use structured outputs (response_format: json_schema) for reliable parsing in automation skills
Known Issues & Gotchas
- No free tier — you must add payment method and fund your account
- Rate limits are tiered based on cumulative spending — new accounts start at Tier 1
- API keys are organization-scoped — be careful with multi-project setups
Frequently Asked Questions
Is OpenAI the best provider for OpenClaw?
It depends on your use case. OpenAI has the broadest model range and best tool-use support. Anthropic's Claude excels at coding and long-context tasks. Google Gemini offers competitive pricing. Most users start with OpenAI and add others as needed.
What's the cheapest way to use OpenAI with OpenClaw?
Use GPT-4o-mini ($0.15/1M input tokens) for routine tasks and reserve GPT-4o for complex reasoning. The mini model handles most agent tasks well at 1/16th the cost of GPT-4o.
Does OpenClaw support all OpenAI models?
Yes. OpenClaw supports GPT-4o, GPT-4o-mini, GPT-4 Turbo, o1, o3-mini, and legacy models through the OpenAI provider configuration. Image generation, TTS, and Whisper are also supported via skills.