Looks like the best option would be OpenAI API using gpt-3.5-turbo.
Cost Example:If a user submits a 500-word post (approximately 667 tokens) and the AI generates a 250-word response (approximately 333 tokens), the costs would be:
- GPT-3.5 Turbo:
- Input: 667 tokens × ($0.0015 / 1,000 tokens) ≈ $0.001
- Output: 333 tokens × ($0.002 / 1,000 tokens) ≈ $0.0007
- Total per fact-check: Approximately $0.0017