O
A
G
x
M
D
CGoogle
See details ↗50% promotional credit
50% billing credit on Gemini 3.6 / 3.7 / 3.8 Flash
AI unit economics
Know what your AI feature costs before the invoice does.
Current AI API pricing across OpenAI, Anthropic, Google and more — plus calculators for cost per request, user, agent and workload.
Latest pricing changes
New models and rates worth checking
Fresh model launches and scheduled price changes, linked directly to cost details and calculators.
Claude Fable 5.2
New successor to Fable 5.1 with lower base and cache-read pricing.
Input / 1M$8
Output / 1M$40
GPT-6.1 Terra
Efficient current-generation model with 1.05M context and 50% Batch/Flex rates.
Input / 1M$1
Output / 1M$5
GPT-6.1 Sol
Near-Astra performance at a lower standard token rate.
Input / 1M$2
Output / 1M$10
Gemini 4 Argon
Introductory standard rates through Nov 30; published pricing doubles Dec 1.
Input / 1M$2
Output / 1M$10
Model intelligence
Current API pricing, across the ecosystem
One representative model from each provider, with current token pricing at a glance. Open a model for context rules and cost details.
OpenAI
GPT-6 AstraInput$10
Output$50
Anthropic
Claude Fable 5.2Input$8
Output$40
Google
Gemini 4 ArgonInput$2
Output$10
xAI
Grok 4.6Input$2
Output$6
Mistral
Mistral Small 4Input$0.15
Output$0.60
DeepSeek
DeepSeek V4.1 FlashInput$0.1938
Output$0.775
Cohere
Command AInput$2.50
Output$10
Calculators
Calculators
Start from the business question you actually have: monthly bill, cost per user, cost per task, or whether a workload fits.
API cost
LLM API Cost Calculator
Estimate monthly API spend from input tokens, output tokens and request volume across major production models.
Calculate → Unit economicsAI Cost per User Calculator
Calculate LLM API cost per active user and compare it with your SaaS subscription price or product margin.
Calculate → Agent costAI Agent Cost Calculator
Estimate AI agent running cost from task volume, multi-step model calls, retries and paid tool usage.
Calculate → RAG costRAG Cost Calculator
Estimate retrieval-augmented generation cost including document embedding, retrieved context, generation and vector database spend.
Calculate → Chatbot costAI Chatbot Cost Calculator
Estimate monthly chatbot API cost using conversation length, history, system prompts and reply size.
Calculate → CachingPrompt Caching Savings Calculator
Estimate how much prompt caching can reduce repeated input-token costs for supported AI models.
Calculate → Batch processingBatch API Savings Calculator
Compare standard LLM processing cost with a configurable batch-processing discount for asynchronous workloads.
Calculate → Context planningLLM Context Window Calculator
Estimate prompt tokens and check whether text plus reserved output fits current model context windows.
Calculate → Traffic planningAI Cost per 1,000 Requests Calculator
Calculate what 1,000 AI API calls cost for a chosen model and average input/output size.
Calculate → Model comparisonAI Model Cost Comparison
Compare modeled monthly API cost across major AI providers and production models using the same workload assumptions.
Calculate →Deals & credits
Credits and promotions worth checking
4 listed offers with official source links. Check eligibility and terms. Not sponsored.
Google for Startups
See details ↗Up to $350K in Cloud credits
Google for Startups AI cloud credits
OpenAI
See details ↗API credits available
OpenAI startup API credits
Anthropic
See details ↗API credits + priority limits
Anthropic Startup Program
Token pricing is easy.
Product economics are not.
A provider's dollars-per-million-tokens table tells you almost nothing about what a feature will cost at production traffic. A real estimate needs a workload shape: how many requests, how much context, how much output and how often.
TokenCOGS converts those moving parts into units product teams can reason about: cost per request, per user, per conversation, per task, per month and per year.
Pricing is synchronized from primary provider pages where supported, with a verified fallback snapshot if the live registry is unavailable. Every result is an estimate and should be verified before procurement or production budgeting.