Forecast AI spend over a growth curve and see when product usage consumes a fixed model budget.
Growth curve
What · The model whose live prices drive this calculation.How · Pick the one you'd actually deploy; each option shows its input/output $/1M rates.Example · Swapping Opus ($5/
5) for Haiku (
/$5) can cut a token bill ~5×.
5.00 out
Gemini 3.5 Flash - Google -
.50 in / $9.00 out
Kimi K2.6 - Moonshot AI - $0.950 in / $4.00 out
MiMo-V2.5-Pro - Xiaomi -
.00 in / $3.00 out
GPT-5.3 Codex - OpenAI -
.75 in /
4.00 out
Grok 4.3 - xAI -
.25 in /
.50 out
Claude Opus 4.6 - Anthropic - $5.00 in /
5.00 out
Qwen3.6 Max Preview - Alibaba -
.30 in / $7.80 out
Claude Sonnet 4.6 - Anthropic - $3.00 in /
5.00 out
DeepSeek V4 Pro - Alibaba (China) - $0.435 in / $0.870 out
GLM-5.1 - Alibaba (China) - $0.870 in / $3.48 out
GPT-5.2 - OpenAI -
.75 in /
4.00 out
GPT-5.2 Chat - OpenAI -
.75 in /
4.00 out
Qwen3.6 Plus - Alibaba - $0.500 in / $3.00 out
GLM-5 - Alibaba (China) - $0.860 in / $3.15 out
Claude Opus 4.5 - Anthropic - $5.00 in /
5.00 out
MiniMax-M2.7 - Alibaba (China) - $0.300 in /
.20 out
MiMo-V2-Pro - Xiaomi -
.00 in / $3.00 out
MiMo-V2.5 - Xiaomi - $0.400 in /
.00 out
GPT-5.2 Codex - OpenAI -
.75 in /
4.00 out
GPT-5.4 mini - OpenAI - $0.750 in / $4.50 out
Gemini 3 Pro Preview - Google -
.00 in /
2.00 out
GPT-5.1 - OpenAI -
.25 in /
0.00 out
GPT-5.1 Chat - OpenAI -
.25 in /
0.00 out
Kimi K2.5 - Moonshot AI - $0.600 in / $3.00 out
GLM-5-Turbo - Z.AI -
.20 in / $4.00 out
DeepSeek V4 Flash - Alibaba (China) - $0.140 in / $0.280 out
Gemini 3 Flash Preview - Google - $0.500 in / $3.00 out
Qwen3.6 27B - Alibaba - $0.600 in / $3.60 out
Qwen3.5 397B-A17B - Alibaba (China) - $0.430 in /
.58 out
GPT-5-Codex - OpenAI -
.25 in /
0.00 out
GPT-5.4 nano - OpenAI - $0.200 in /
.25 out
Qwen3.6 35B-A3B - Alibaba - $0.248 in /
.49 out
MiMo-V2-Omni - Xiaomi - $0.400 in /
.00 out
GPT-5.1 Codex - OpenAI -
.25 in /
0.00 out
GLM-5V-Turbo - Z.AI -
.20 in / $4.00 out
Qwen3.5 27B - Alibaba - $0.300 in /
.40 out
GLM-4.7 - Vertex - $0.600 in /
.20 out
MiniMax-M2.5 - Alibaba (China) - $0.300 in /
.20 out
DeepSeek V3.2 - Vertex - $0.560 in /
.68 out
Qwen3.5 122B-A10B - Alibaba - $0.400 in / $3.20 out
MiMo-V2-Flash - Xiaomi - $0.100 in / $0.300 out
Kimi K2 Thinking - Vertex - $0.600 in /
.50 out
MiniMax-M2.1 - MiniMax (minimax.io) - $0.300 in /
.20 out
Mistral Medium 3.5 - Mistral -
.50 in / $7.50 out
GPT-5.1 Codex mini - OpenAI - $0.250 in /
.00 out
Step 3.5 Flash 2603 - StepFun (China) - $0.100 in / $0.300 out
Step 3.5 Flash - StepFun (China) - $0.096 in / $0.288 out
Qwen3.5 35B-A3B - Alibaba - $0.250 in /
.00 out
MiniMax-M2 - MiniMax (minimax.io) - $0.300 in /
.20 out
Gemini 3.1 Flash Lite - Google - $0.250 in /
.50 out
Gemini 3.1 Flash Lite Preview - Google - $0.250 in /
.50 out
Mercury 2 - Inception - $0.250 in / $0.750 out
GLM-4.6 - Z.AI - $0.600 in /
.20 out
Qwen3 Max - Alibaba (China) - $0.861 in / $3.44 out
Kimi K2 0905 - Moonshot AI - $0.600 in /
.50 out
Mistral Small 4 - Mistral - $0.150 in / $0.600 out
GLM-4.6V - Z.AI - $0.300 in / $0.900 out
Devstral 2 - Mistral - $0.400 in /
.00 out
Mistral Small (latest) - Mistral - $0.150 in / $0.600 out
Mistral Medium (latest) - Mistral -
.50 in / $7.50 out
Qwen3-Omni Flash - Alibaba (China) - $0.058 in / $0.230 out
GLM-4.7-FlashX - Z.AI - $0.070 in / $0.400 out
DeepSeek Chat - DeepSeek - $0.140 in / $0.280 out
DeepSeek Reasoner - DeepSeek - $0.140 in / $0.280 out
Gemini Flash-Lite Latest - Google - $0.100 in / $0.400 out
solar-pro3 - Upstage - $0.250 in / $0.250 out
siliconflow/deepseek-v3.2 - Alibaba (China) - $0.270 in / $0.420 out
Mercury Edit 2 - Inception - $0.250 in / $0.750 out
Qwen3.6 Flash - Alibaba - $0.188 in /
.13 out
siliconflow/deepseek-v3.1-terminus - Alibaba (China) - $0.270 in /
.00 out
Qwen3-VL Plus - Alibaba (China) - $0.143 in /
.43 out
MiniMax-M3 - MiniMax (minimax.io) - $0.300 in /
.20 out
Qwen3.5 Flash - Alibaba (China) - $0.172 in /
.72 out
Gemini Flash Latest - Google - $0.300 in /
.50 out
Qwen3.5 Plus - Alibaba - $0.400 in /
.40 out
Moonshot Kimi K2 Thinking - Alibaba (China) - $0.574 in /
.29 out
Moonshot Kimi K2.5 - Alibaba (China) - $0.574 in /
.41 out
MiniMax-M2.5-highspeed - MiniMax (minimax.io) - $0.600 in /
.40 out
MiniMax-M2.7-highspeed - MiniMax (minimax.io) - $0.600 in /
.40 out
Grok Build 0.1 - xAI -
.00 in /
.00 out
kimi/kimi-k2.5 - Alibaba (China) - $0.600 in / $3.00 out
Grok 4.20 (Non-Reasoning) - xAI -
.25 in /
.50 out
Grok 4.20 Multi-Agent - xAI -
.25 in /
.50 out
Moonshot Kimi K2.6 - Alibaba (China) - $0.929 in / $3.86 out
Claude Haiku 4.5 - Anthropic -
.00 in / $5.00 out
Kimi K2 Thinking Turbo - Moonshot AI -
.15 in / $8.00 out
GPT-5.1 Codex Max - OpenAI -
.25 in /
0.00 out
Kimi K2 Turbo - Moonshot AI -
.40 in /
0.00 out
Gemini 3.1 Pro Preview Custom Tools - Google -
.00 in /
2.00 out
GPT-5.3 Chat (latest) - OpenAI -
.75 in /
4.00 out
GPT-5.3 Codex Spark - OpenAI -
.75 in /
4.00 out
Claude Sonnet 4.5 - Anthropic - $3.00 in /
5.00 out
Nano Banana 2 - Google - $0.500 in / $60.00 out
GPT-5 Pro - OpenAI -
5.00 in /
20.00 out
GPT-5.2 Pro - OpenAI -
1.00 in /
68.00 out
GPT-5.4 Pro - OpenAI - $30.00 in /
80.00 out
GPT-5.5 Pro - OpenAI - $30.00 in /
80.00 out
What · The total AI spend you're willing to burn over the projection window.How · Sets the runway line: the month your projected cumulative cost crosses it.Example · A $50,000 budget might cover 8 months before fast user growth blows past it.
What · Your active-user count in month 1, before growth is applied.How · Monthly growth compounds on top of this to project a 12-month cost curve.Example · 4,000 starting users at 18%/month becomes ~28,000 — and ~7× the bill — by month 12.4,000
What · Percentage your active-user base grows each month.How · Compounds monthly, so cost rises faster than linearly; match it to your real adoption.Example · 18%/month roughly doubles your users — and AI spend — every 4–5 months.18%
What · Average model calls each active user triggers per day.How · Total volume = users × this × work days, so it scales your whole bill.Example · 1,000 users sending 3 chat messages/day = 3,000 calls/day.1.4
What · The tokens you send the model each call — your prompt, system instructions, and any attached context.How · Raise it to model longer prompts or more context; it's billed at the model's input rate and usually drives most of the bill on long prompts.Example · 6,000 input tokens ≈ a 4–5 page document pasted into ChatGPT; double it and your input cost doubles.1,800
What · The tokens the model generates back each call — how long the answer is.How · Higher values allow longer, more detailed replies, but output is billed 3–5× the input rate, so these are the priciest tokens.Example · Like ChatGPT's 'Maximum length': 4,000 output tokens ≈ a multi-page report; 400 ≈ a short answer.420