Question 1

How much does the GPT-5.4, Claude Opus or Gemini 3.1 API cost per 1M tokens?

Accepted Answer

Prices per million tokens change frequently. Use the table above for the latest available rate for each model: the catalog syncs with official sources and shows when each value was verified. Change the currency selector to see them in EUR or any other currency.

Question 2

What is a token and how many tokens does a word have?

Accepted Answer

A token is the unit an AI model processes: usually between half a word and a full word. Rule of thumb: 1,000 tokens ≈ 750 English words. A 20-word sentence is about 26 tokens; a 300-word email is around 400. Models charge for input tokens (your prompt) and output tokens (their answer) separately.

Question 3

What is a "context window" and why does it matter?

Accepted Answer

The context window is the maximum number of tokens a model can process in a single call (prompt + answer). 200,000 tokens fit ~2 long PDFs; 1M fits ~10 full novels. If you're analysing contracts or large codebases you need large windows. For a support chatbot, 32K is usually plenty.

Question 4

Why do cached tokens cost up to 90% less?

Accepted Answer

When your app sends the same system prompt on every request (e.g. "You are a customer support agent for Acme…"), providers store it once and reuse the computation. They only charge full price for the new part (the user's question). OpenAI, Anthropic and Google give 50% to 90% off cached tokens. The calculator above accounts for this via a cache-hit % slider.

Question 5

What's the cheapest AI model in 2026?

Accepted Answer

For general use, DeepSeek V3 ($0.014 input / $0.028 output per 1M) and Llama 4 Scout ($0.08 / $0.30) are the cheapest in the table. For complex reasoning, Qwen and DeepSeek R1 are the most competitive against o3 or GPT-5.4. Filter by "Budget" category in the comparator to see the current ranking.

Question 6

Which model should I use for a chatbot, code generation or document analysis?

Accepted Answer

Chatbot: GPT-5.4 Mini or Claude Haiku 4.5 for latency; DeepSeek V3.2 for cost. Code: Claude Opus 4.6 and GPT-5.4 still lead the benchmarks, with o3 and DeepSeek R1 for hard reasoning. Document analysis: you need a large window — Gemini 3.1 Pro, Claude Opus or Llama 4 Maverick. Use the Model Finder above for a personalised pick.

Question 7

How does GPT-5.4 differ from Claude Opus 4.6?

Accepted Answer

GPT-5.4 is faster, half the input price ($2.50 vs $5 per 1M) and has better function-calling and batch API ergonomics. Claude Opus 4.6 wins at long-form writing, large codebases and strict multi-step instructions. Both ship 1M-token context windows. If cost is no issue: Claude for long text and craft; GPT-5.4 for latency and volume.

Question 8

How much will I spend per month for N AI requests?

Accepted Answer

Open the calculator above, pick your use cases (emails, PDFs, chats…) or enter tokens manually, and you'll see cost per request, per day and per month in your chosen currency. For reference: a 1,000 requests/day chatbot on GPT-5.4 Mini costs ~$60/mo; on DeepSeek V3, ~$5/mo. The calculator also sums a combined total when you compare several models.

Question 9

Are the prices in this tool up to date?

Accepted Answer

The catalog checks its configured official sources several times a day. The status and verification date appear next to the comparator; if a provider is unavailable, we keep the last valid snapshot and mark it as stale.

Question 10

What's the difference between flagship, mid-tier, budget and reasoning models?

Accepted Answer

Flagship: the provider's top model (GPT-5.4, Claude Opus 4.6, Gemini 3.1 Pro). Mid-tier: a trimmed version with ~80% of the quality at ~30% of the cost (GPT-5.4 Mini, Claude Sonnet). Budget: optimised for massive volume (GPT-5.4 Nano, Claude Haiku, DeepSeek V3). Reasoning: trained to think step-by-step — slower but better at hard problems (o3, DeepSeek R1, Sonar Reasoning).

Question 11

Which provider has the largest context window?

Accepted Answer

The model with the largest context window can change as providers release new versions. Use the "Context Windows" tool above to see the current ranking and check how many tokens your content weighs.

Question 12

Are there open-source models cheaper than GPT?

Accepted Answer

Yes. Llama 4 (Meta), DeepSeek V3, Qwen 3.5 and Mistral are open or open-weight and can be self-hosted. DeepSeek V3 on API costs about 1% of GPT-5.4 with comparable quality in many cases. If you handle sensitive data, an on-prem or VPC deployment with Llama 4 Maverick can beat any hosted API above a certain volume.

Question 13

How do I cut AI costs in my application?

Accepted Answer

Five levers: (1) cache system prompts (saves 50-90%), (2) batch API for non-urgent requests (50% off), (3) smaller model for simple tasks and flagship only when needed (complexity router), (4) trim prompts and cap output tokens, (5) fine-tune or use RAG instead of re-sending the same context every call. At Letbrand we audit each lever and typically cut costs by 40-80%.

Question 14

What is a "reasoning model" (o3, DeepSeek R1)?

Accepted Answer

They're models trained to think step by step before answering, generating an internal chain of reasoning. Slower and pricier per token than standard models, but much better at math, logic and hard code. Use them selectively: not worth it for a customer-support bot, but great for technical analysis, debugging or autonomous agents.

Question 15

How do you at Letbrand pick the model for a client?

Accepted Answer

We map the use case against three axes: required quality, acceptable latency and monthly cost. We A/B 2-3 candidates on a real client dataset and measure accuracy + cost + latency. Then we wire caching, complexity routing and observability. The result is typically 40-80% cheaper than the obvious first pick. Want us to do it for you? Let's talk.

Browse AI models

Token Price Comparator

Frequently asked questions about AI models and pricing

How much does the GPT-5.4, Claude Opus or Gemini 3.1 API cost per 1M tokens?

What is a token and how many tokens does a word have?

What is a "context window" and why does it matter?

Why do cached tokens cost up to 90% less?

What's the cheapest AI model in 2026?

Which model should I use for a chatbot, code generation or document analysis?

How does GPT-5.4 differ from Claude Opus 4.6?

How much will I spend per month for N AI requests?

Are the prices in this tool up to date?

What's the difference between flagship, mid-tier, budget and reasoning models?

Which provider has the largest context window?

Are there open-source models cheaper than GPT?

How do I cut AI costs in my application?

What is a "reasoning model" (o3, DeepSeek R1)?

How do you at Letbrand pick the model for a client?

Token Price Comparator

Frequently asked questions about AI models and pricing

How much does the GPT-5.4, Claude Opus or Gemini 3.1 API cost per 1M tokens?

What is a token and how many tokens does a word have?

What is a "context window" and why does it matter?

Why do cached tokens cost up to 90% less?

What's the cheapest AI model in 2026?

Which model should I use for a chatbot, code generation or document analysis?

How does GPT-5.4 differ from Claude Opus 4.6?

How much will I spend per month for N AI requests?

Are the prices in this tool up to date?

What's the difference between flagship, mid-tier, budget and reasoning models?

Which provider has the largest context window?

Are there open-source models cheaper than GPT?

How do I cut AI costs in my application?

What is a "reasoning model" (o3, DeepSeek R1)?

How do you at Letbrand pick the model for a client?