High concurrency, short output. Prioritize low input + output price.
๐ป
Scenario B: Code / Agent
Medium-long output. Focus on output price + code quality.
๐ง
Scenario C: Reasoning / Math
Long chain reasoning. Look for reasoning capability + long context.
๐ Price Visualization
Visual comparison of model costs
Cost per 1M Tokens (Input + Output)
Price vs Context Length
If you only need short conversations, you don't need the most powerful model. If you frequently generate long outputs, output price is your main cost driver.
๐ก Understanding Token Costs
Tokens are billing units, not equal to words or characters
English: 1 token โ 4 characters โ 0.75 words
Chinese: 1 token โ 1-2 characters (varies by model)
It depends on your use case. For short conversations, look at input price. For long outputs, focus on output price. Currently, DeepSeek V3 and Gemini Flash series offer the best value for most use cases.
Use the same metric: $/1M tokens. Compare input and output prices separately, then use the calculator with your actual input/output ratio to estimate real costs.
Use the calculator's text estimation mode. As a rough guide: English 1 token โ 4 characters โ 0.75 words. Chinese 1 token โ 1-2 characters.
Most coding tasks generate longer outputs, so output price usually has more impact on total cost. Consider weighting output price higher in your comparison.
Reasoning models (like o1, DeepSeek R1) typically generate longer thinking chains and output tokens. Consider controlling output length or using cheaper draft models before switching to powerful ones.
We regularly update pricing data from official sources. Check the 'Updated' timestamp at the bottom of each page. For critical decisions, always verify with official pricing pages.
Enter your token usage in the calculator. The system will suggest models with similar context length and capabilities at lower prices.
Different channels (official API vs cloud providers) may have different pricing, billing methods, or regional fees. We label the provider_channel field to distinguish these.
