Agent Cost Calculator

Model the real cost of running an AI agent — before you scale. Adjust inputs below to see cost per run, per day, and per month.

Model

16 of 16

Tier

Type

Strength

Token Profile

System prompt / context2,000 tokens

Usually cached — docs, instructions, retrieved context

Input per run1,500 tokens

User message + any per-run dynamic context

Output per run500 tokens

Tool Calls

Tool calls per run3 calls

Each call adds an extra input + output round

Tokens per tool call300 tokens avg

Prompt Caching

Cache hit rate70%

How often the system prompt is served from cache

Volume

Runs per day100

Cost Estimate

Per run$0.0217

Per day$2.173

Per month$65.20

Breakdown

Uncached input$0.0032 (14%)

Cached reads$0.7350m (3%)

Cache writes$0.0023 (10%)

Output$0.0075 (35%)

Tool calls$0.0081 (37%)

Prices verified May 2026 — verify against provider docs before scaling.

Comparison

same config · cheapest first

DeepSeek V4 Flashopen$0.4829m/run

Llama 3.3 70Bopen$0.6990m/run

Grok 4.1 Fast$0.8975m/run

DeepSeek V4 Proopen$0.0015/run

Mistral Large 2open$0.0023/run

MiniMax M2.7open$0.0023/run

Qwen 3.6 Plusopen$0.0031/run

Gemini 3 Flash$0.0037/run

Kimi K2.6open$0.0048/run

GPT-5.4 mini$0.0056/run

GLM 5.1open$0.0062/run

Claude Haiku 4.5$0.0072/run

Gemini 3.1 Pro$0.0149/run

Claude Sonnet 4.6$0.0217/run

Claude Opus 4.7$0.0362/run

GPT-5.5$0.0372/run

Built by Sai Viki · Pricing via OpenRouter + provider docs · Lineup curated from real-usage rankings