CostForge
AI API Cost Calculator
Compare Claude vs GPT vs Gemini API costs side-by-side. Calculate daily and monthly spend, find the cheapest model for your use case. Free, no sign-up.
📝 Token Estimator
💰 Prompt Caching Savings
💡 Cost Optimization Tips
Why CostForge?
20 Models Compared
Side-by-side pricing for Claude, GPT-5.5, Gemini, DeepSeek V4, Grok 4.3, and Mistral — 6 providers in one view.
Real-Time Calculations
Drag sliders or type exact numbers. See per-request, daily, and monthly costs update instantly across all models.
6 Use Case Presets
Pre-configured token counts for Chatbot, Code Assistant, Data Processing, RAG Pipeline, Content Generation, and Summarization.
Batch vs Standard
Toggle between standard and batch pricing. See exactly how much you save with async/batch processing for each provider.
Visual Cost Bars
Horizontal bars make it easy to compare relative costs at a glance. The cheapest model is highlighted with a BEST badge.
Optimization Tips
Expert advice on reducing API costs: prompt caching, model routing, token compression, and budget monitoring strategies.
Always Up-to-Date
Pricing data is based on latest published rates from Anthropic, OpenAI, Google, DeepSeek, xAI, and Mistral as of May 2026.
100% Private
All calculations happen in your browser. No data is sent anywhere. No account needed, no tracking.
Frequently Asked Questions
How accurate are these prices?
Prices are based on the latest published pricing from Anthropic, OpenAI, Google, DeepSeek, xAI, and Mistral as of May 2026. We update regularly, but always verify against the official pricing pages for the most current rates. Actual costs may vary with prompt caching, commitments, or volume discounts.
What's the difference between standard and batch pricing?
Batch/async pricing is typically 50% cheaper but has higher latency (responses delivered within hours, not seconds). Use batch for non-real-time tasks like data processing, analysis, and bulk content generation.
How do I estimate my token usage?
A rough rule: 1 token ≈ 4 characters in English. A typical user message is 50-500 tokens, a code file is 1,000-5,000 tokens, and a long document is 10,000+ tokens. Output tokens are usually 20-50% of input tokens for most tasks.
Which model should I choose?
For simple Q&A and classification: use DeepSeek V4-Flash/Gemini Flash-Lite/Mistral Small (cheapest). For coding and analysis: use Sonnet 4.6/GPT-4.1/Grok 4.3 (best value). For complex reasoning: use Opus 4.7/GPT-5.5/o3/Gemini Pro (most capable). Always benchmark on your specific use case.
Does this include hidden costs like network, storage, etc.?
No. CostForge calculates pure API token costs only — what the providers charge per input/output token. Other costs like network bandwidth, compute for pre/post-processing, and database storage are separate and depend on your infrastructure.