AI API cost calculator
Monthly cost for 15 APIs, cheapest first. Prices verified 2026-10-17, each with its source.
| API and plan | Per month | Details |
|---|---|---|
| OpenAI Moderationomni-moderation-latest · Free | $0 | |
| Perplexity Decisions APIpplx-decider-flash-v1.2 · $0.01 in / $0 out per 1M tokens | $1.00 | |
| Inception Mercury Decidemercury-decide (promo) · $0.02 in / $0 out per 1M tokens | $2.00 | |
| Liquid AI d1d1 · $0.02 in / $0 out per 1M tokens | $2.00 | |
| Perplexity Decisions APIpplx-decider-v1.2 · $0.02 in / $0 out per 1M tokens | $2.00 | |
| Cloudflare ClefClef-flash · $0.038 in / $0 out per 1M tokens | $3.80 | |
| Jev (TypeSafe AI)jev-1.13 / jev-1.14 · $0.042 in / $0 out per 1M tokens | $4.20 | |
| Liquid AI d1-ultrad1-ultra · $0.042 in / $0 out per 1M tokens | $4.20 | |
| LLM Gateway — DecisionsJev 1.13 via LLM Gateway · $0.042 in / $0 out per 1M tokens | $4.20 | |
| Microsoft-Decision-1via OpenRouter (provider Azure) · $0.042 in / $0 out per 1M tokens | $4.20 | |
| AWS Bedrock GuardrailsInvokeGuardrailChecks content filters · $0.07 per 1,000 text units | $7.00 | |
| OpenAI Decisions APIgpt-6-luna · $0.1 in / $0 out per 1M tokens | $10.00 | |
| AWS Bedrock GuardrailsContent filters (text) · $0.15 per 1,000 text units | $15.00 | |
| Cloudflare ClefClef-omni · $0.15 in / $0 out per 1M tokens | $15.00 | |
| Cloudflare ClefClef · $0.24 in / $0 out per 1M tokens | $24.00 | |
| Anthropic Claude (structured outputs)Haiku 5.5 · $0.1 in / $0.5 out per 1M tokens | $25.00 | |
| Azure AI Content SafetyS0 text records (East US) · $0.375 per 1,000 text units | $37.50 | |
| DeepSeek Search & Fetch APISearch · $0.75 per 1,000 calls | $75.00 | |
| DeepSeek Search & Fetch APIFetch · $0.75 per 1,000 calls | $75.00 | |
| Gemini structured output (Vertex AI)Gemini 3.8 Flash (introductory) · $0.75 in / $3.75 out per 1M tokens | $188 | |
| Anthropic Claude (structured outputs)Haiku 4.5 · $1 in / $5 out per 1M tokens | $250 | |
| Anthropic Claude (structured outputs)Sonnet 5.5 · $2 in / $10 out per 1M tokens | $500 | |
| Anthropic Claude (structured outputs)Opus 5.5 · $4 in / $20 out per 1M tokens | $1,000 |
List prices before free allowances and taxes. Formula on hover (cost cell) or under Formula. Presets are our assumptions, not vendor figures.
Not shown for this usage: AWS Bedrock Guardrails (Content filters (image)); Azure AI Content Safety (S0 images (East US)). Per-image and per-text-unit plans appear when those inputs are above 0.
Not comparable
| OpenRouter Armor | Free during the beta; OpenRouter says GA pricing will be announced later, and your normal model inference is billed separately. |
|---|---|
| Vercel AI Gateway — Decisions | Pass-through: no markup, you pay the routed provider’s list price, so the cost depends on which model you route to. |
| Strands Decider 2B | Open weights; you pay your own compute, which depends on your hardware. No per-request price exists. |
| Laya | Open weights; you pay your own compute, which depends on your hardware. No per-request price exists. |
| llama.cpp decision endpoint | Free open-source server; compute cost depends on your hardware. |
| H2O-Lightning-4B | Open weights; the $0.021 per 1,000 decisions figure is H2O’s own self-hosting estimate, not a price. |
| Cohere Classify | No Classify price is published on cohere.com/pricing. |
Use this from your agentHTTP API and MCP
Agents can call the same pricing engine. No key, CORS open, JSON out. Full reference on For agents and in openapi.json. Listed in the directory as our MCP server and our cost API.
Estimate cost over HTTP
curl "https://agentapilist.com/api/cost?requests=100000&input_tokens=1000&output_tokens=300&category=decision"
Remote MCP server (Streamable HTTP)
{
"mcpServers": {
"agentapilist": { "type": "http", "url": "https://agentapilist.com/mcp" }
}
}Units and presets
| tokens_1m | USD per 1M input / output tokens |
| text_units_1k | USD per 1,000 text units (up to 1,000 characters each) |
| per_image | USD per image |
| requests_1k | USD per 1,000 calls (each request is one call; tokens do not matter) |
| free | Free (no charge per request) |
| Preset | Input | Output | Text units |
|---|---|---|---|
| Classification | 500 | 5 | 1 |
| Chatbot turn | 1,000 | 300 | 1 |
| Agent loop step | 4,000 | 500 | 4 |
Presets are our assumptions for typical requests, not vendor figures. Change any input to match your workload.
Cost by APIPre-computed tables
- OpenAI Decisions API cost
- Microsoft-Decision-1 cost
- Jev cost
- Cloudflare Clef cost
- Perplexity Decisions cost
- Mercury Decide cost
- Liquid d1 cost
- Liquid d1-ultra cost
- LLM Gateway Decisions cost
- Bedrock Guardrails cost
- Azure Content Safety cost
- Gemini 3.8 Flash cost
- Claude API cost
Head to head
- OpenAI Decisions API vs Jev
- Microsoft-Decision-1 vs Jev
- OpenAI Decisions API vs Microsoft-Decision-1
- Jev vs Cloudflare Clef
- OpenAI Decisions API vs Cloudflare Clef
- Bedrock Guardrails vs Azure Content Safety
- Gemini 3.8 Flash vs Claude API
- Perplexity Decisions vs Liquid d1
- Mercury Decide vs Perplexity Decisions
- Liquid d1-ultra vs Perplexity Decisions
FAQ
How is the monthly cost calculated?
For token-priced APIs: requests × (input tokens × input price + output tokens × output price) ÷ 1,000,000. For AWS Bedrock Guardrails and Azure Content Safety: requests × text units × price per 1,000 text units. Per-image plans multiply requests × images × price per image.
Are free tiers deducted?
No. Costs are list prices before free allowances and taxes. Each row notes its free tier, for example Azure F0’s 5,000 free text records a month or Cloudflare’s 10,000 free Neurons a day.
Where do the prices come from?
Each price was verified on the vendor’s pricing page and shows its date and source link. The table was last checked 2026-10-17.
Why are some APIs missing?
7 APIs are listed as not comparable: open-weight models you self-host, a pass-through gateway, a free beta with no announced price, and one API with no published price.