AI API Cost Calculator
What does your AI feature actually cost?
Pick a model. Set your workload. See daily, monthly, and annual cost - with the real optimizations most teams miss.
Pricing verified: 2026-07-28
163 models across 16 providers
Caching + batch API applied
What this calculator does
See exactly what an LLM workload will cost across 70+ models. Pick a model, enter your tokens per request and daily volume, get per-request / daily / monthly / annual cost. Caching and Batch API savings calculated automatically.
Why use it
- Stop guessing — turn "AI is expensive" into a precise monthly number you can defend to finance
- Compare 70+ models side-by-side at YOUR token shape, not vendor marketing examples
- Spot the 30-90% savings opportunities (prompt caching, Batch API, model swap) before you ship
- Re-cost instantly when a vendor changes rates — your numbers stay current
New to this calculator? Start with the ⚡ Playground — a few sliders, instant ballpark. Then switch to the 🧮 Calculator for your exact number.
Two ways to use this: visualize in the Playground, then get your number in the Calculator.
What will this cost per month?
The three inputs that move the bill. We assume a balanced-tier model and ~30 working days.
Estimated monthly cost
—
Change the input sliders below to see new estimates.
How we got this estimate
—
💡Cost = (input + output tokens) × the model’s rate × requests × ~30 days. Switch model tiers in the full calculator.
📊 Not sure of a value? Fields with a ▾ Typical pill offer broad industry ballparks (sourced typical ranges) so you can move forward now — your result gets more accurate as you replace them with your own measured numbers. Values marked * are rough estimates.
📥 Inputs you provide
- ModelPick from 70+ AI models
- Input tokens per requestSize of your prompt
- Output tokens per requestExpected response size
- Requests per dayYour daily call volume
- Prompt cache hit rateHow often your prompt prefix repeats
- Days per monthWorking days for billing math
📤 Outputs you get
- Cost per requestDollars per single API call
- Monthly costDollars per month at your volume
- Annual costLinear annual projection
- Input vs output cost splitWhere the money goes
- Optimization suggestionsHow to cut the bill
🎯 Use your results to
🎯
Pick the right model
Run the same workload through 5 candidates; pick the cheapest that meets your quality bar
📈
Forecast your AI bill
Defensible monthly + annual numbers for your finance team
💾
Quantify savings
Estimated dollars from caching, Batch API, and model swap — before you implement
🔌
Integrate with your AI agents
MCP available for agentic workflow integration — surface live cost intelligence to your agents
1
Enter key values
Input tokens, output tokens, and pick a model. That's it.
2
Compare models
Scroll the model list — cheapest first. All priced at your exact workload.
3
Read per-request cost
See the per-call dollar amount and multiply by your expected monthly volume.
1
Enter every field
Input tokens = everything you send (system prompt + user message + retrieved docs + examples). Output tokens = what the model generates. Rough rule: 1 token ≈ 0.75 English words. If you don't know your exact counts, use Token Estimator to measure real samples first.
2
Choose a model
Models listed with input / output prices per 1M tokens. Frontier work → Claude Opus 4.8 (newest, May 2026), Gemini 3.1 Pro, or GPT-5.5. Production traffic → Claude Sonnet 4.6 or Gemini 3 Flash. High-volume cheap → Claude Haiku 4.5, GPT-5-nano, or DeepSeek V3.2. Cheaper isn't always better: slower/weaker models may produce longer outputs, eroding the price advantage.
3
Interpret results
Shows cost per single request, daily, monthly, and annual. Compare columns to find the cost-optimal model at your workload shape. If output share > 70% of cost, you're probably over-generating.
4
Go deeper
This is a baseline — real savings come from routing mixed complexity, caching static prefixes (90% off cached reads at Anthropic/OpenAI/Google), and batching async work (50% off at Anthropic/OpenAI/Google Vertex).
👇 Now try the calculator below with your own AI workloads
⚙️ Calculator inputs — tap to edit your workload
🎛 CALCULATOR
Your workload
Estimate conservatively - we'll show you what caching + batch mode save below.
Load a typical workload, then tweak the numbers.
Hint: Pick the model you'll run. The note below shows its rates.
Hint: Words you send in: prompt + question + docs. Short ~250, doc-heavy a few thousand.
Hint: Words it sends back. A line ~40, a paragraph ~200, long ~1,000+.
Hint: Uses per day at your busiest. Multiplies everything.
Hint: How often it re-reads the same setup text. High for chatbots; 0 if unsure.
Hint: 30 for everyday apps; 22 for business-days-only.
📈 RESULTS
💰 Your estimated cost
Loading…
Monthly cost
-
-
Per request-
Per day-
Input tokens/day-
Output tokens/day-
Input cost share
Output cost share
Annual-
Monthly tokens-
📋 What now?
- Compare models — switch the model dropdown to see the same workload across 70+ options
- Lock in savings — toggle caching and Batch mode to surface the 30-90% reductions before you ship
- Set your budget — use the monthly + annual numbers as defensible inputs for finance
Need help cutting your AI bill?
💼 Talk to a CloudIntelligence advisor →
Now that you have your number…
What this means + what to do next
💡 What to consider beyond this number for full TCO
- Observability + logging (prompts, outputs, latency, errors) — typically adds 5-10% to inference cost at production scale
- Eval pipelines + benchmark sets — $500-$5K/mo even without continuous evaluation; budget more if quality drift matters
- Human-in-the-loop review for edge cases — $4K-$12K/mo per FTE reviewer for production AI features
- Retry / fallback overhead — typically 3-15% on top of base inference depending on error rate and retry logic
- Vendor lock-in cost — invisible until migration day, often $50K+ in re-prompting + re-eval + downtime risk
Rule of thumb: Multiply this number by 1.5–2.5× for production-ready TCO. Lower end (1.5×) = internal tools with low error tolerance and no compliance overhead. Higher end (2.5×) = customer-facing AI features with eval pipelines, compliance logging, and human review.
Quantify the hidden costs:
- If your workload is multi-turn (chat, agents, tool-using), costs compound per turn — this baseline misses that Agent Loop Cost
- Quantifies lock-in cost on the day you need to switch vendors Vendor Concentration Risk
- If you're adding retrieval, the embedding + vector DB + rerank costs aren't in this baseline Rag Pipeline
$ How this fits your overall ROI
This calculator gives you the cost number. Here's how to turn that into an ROI story:
- What revenue or cost-saved does this AI feature drive monthly?
- How long until cumulative AI cost exceeds the value the feature generates?
- How sensitive is your business to vendor price changes? (Last 12 months saw -50% to +25% swings across major vendors.)
Bridge to ROI:
- Convert per-request cost into per-customer or per-feature margin Margin Calculator
- Project 12 months out with growth + price-change assumptions Annual Cost Forecaster
- See cost at 10× and 100× current usage — the discontinuities matter Scale Projection
Doing something different?
Doing something different? These calculators may fit better:
- For multi-turn agent loops with tool calls Agent Loop Cost
- For full RAG over a knowledge base with embeddings + retrieval Rag Pipeline
- For image / multimodal workloads where pricing differs Vision Cost