I spent time compiling pricing data for 50+ LLM models across OpenAI, Anthropic, Google, Mistral, and others. Here's what I found.
The Price Range is Wild
- Cheapest: Gemini 1.5 Flash 8B at $0.19/1M tokens
- Most expensive: o1 Pro at $750/1M tokens
- That's a 3,947x difference
Quick Comparisons
Frontier models (1M tokens):
| Model | Input | Output | Total |
|---|---|---|---|
| GPT-5 | $1.25 | $10.00 | $11.25 |
| Claude Opus 4.5 | $5.00 | $25.00 | $30.00 |
| Gemini 2.5 Pro | $1.25 | $10.00 | $11.25 |
Budget-friendly options:
| Model | Input | Output | Total |
|---|---|---|---|
| GPT-5-mini | $0.25 | $2.00 | $2.25 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $6.00 |
| Gemini 2.0 Flash | $0.10 | $0.40 | $0.50 |
Key Takeaways
- Output tokens cost 3-8x more than input - optimize your max_tokens
- Newer isn't always pricier - GPT-5 is cheaper than GPT-4 Turbo
- "Mini" models are underrated - GPT-5-mini costs 80% less than GPT-
- Google is aggressive on pricing - Gemini 2.0 Flash at $0.50/1M is hard to beat.
Full breakdown with all 50+ models:https://withorbit.io/blog/llm-pricing-comparison-50-models
What models are you using? Curious how others are balancing cost vs capability.
Top comments (0)