#tHow Many Tokens?

← All models

GPT-4o: token counter & pricing

OpenAI · exact (uses official tokenizer) · pricing as of 2026-05-31.

Updated · By · Methodology

Provider
OpenAI
API model ID
gpt-4o-2024-08-06
Context window
128,000 tokens
Input price
$2.50 per 1M tokens
Output price
$10.00 per 1M tokens
Tokenizer accuracy
exact (uses official tokenizer)
Pricing as of
2026-05-31

Open the counter to count tokens for GPT-4o in real time.

What is GPT-4o?

GPT-4o ("o" for omni) is OpenAI's flagship general-purpose model, the workhorse for chat, coding, RAG, and most production AI workloads. Faster and cheaper than the older GPT-4 Turbo, with comparable or better quality on most benchmarks.

How tokens are counted here

GPT-4o introduced the o200k_base tokenizer that the entire GPT-5 family later inherited. We run that exact encoding in your browser with js-tiktoken: your prompt is never uploaded anywhere, and the result is exact rather than estimated. One caveat specific to GPT-4o: this counter measures text tokens only. Images you send to GPT-4o are billed separately by the API based on resolution and detail level, so a vision-heavy workload will cost more than the text count alone suggests.

Pricing notes

OpenAI charges separately for input and output tokens, with output ~4× the input rate. The "Per call" column above assumes the input/output split you set with the slider. The default 80/20 split reflects typical chat workloads where the prompt and history are large but the model's reply is short.

When to use GPT-4o over GPT-4o mini

For high-volume classification, extraction, or short Q&A, GPT-4o mini is ~17× cheaper with a small quality gap that's invisible for most workloads.

What GPT-4o costs in production

A workload GPT-4o still handles well is multimodal document processing: invoices, scanned forms, contracts with embedded figures. Say each document averages 3,000 input tokens of extracted text and instructions, returns a 500-token structured result, and you process 200,000 documents per month. That is 600M input tokens and 100M output tokens.

The same volume on GPT-4o mini ($0.15 / $0.60) is $90 + $60 = $150 per month, a 94% cut for documents simple enough that retrieval and OCR did the hard part. Stepping up to Claude Sonnet 4.6 ($3.00 / $15.00) costs $1,800 + $1,500 = $3,300 per month. Route by document difficulty and the blended bill lands well under either extreme.

Migrating from GPT-4 Turbo

Change the apiId from gpt-4-turbo-2024-04-09 to gpt-4o-2024-08-06. Unlike most upgrades in this family, the tokenizer changes too: GPT-4 Turbo used cl100k_base, GPT-4o uses o200k_base, which produces roughly 5-10% fewer tokens for the same English text. Re-measure your prompt budgets rather than carrying old counts forward. The pricing move is large: $10.00 / $30.00 drops to $2.50 / $10.00 per 1M tokens, a 75% input cut. And once you are on o200k_base, counts do carry over to every GPT-5-era model, so this migration is the last re-measurement you should need.

GPT-4o vs the obvious alternative

Claude Sonnet 4.6 is the perennial head-to-head: $3.00 / $15.00 against GPT-4o's $2.50 / $10.00, so GPT-4o is cheaper per token on both sides, while Sonnet tends to win on instruction-following nuance. The full breakdown lives at GPT-4o vs Claude Sonnet. If your real question is whether you need this tier at all, see GPT-4o vs GPT-4o mini; for the premium direction, Claude Opus vs GPT-4o.

Common questions

Is the o200k tokenizer the same as cl100k (used by GPT-4 Turbo)?

No. o200k_base has a larger vocabulary (~200,000 tokens vs ~100,000) and produces fewer tokens for the same English text, usually 5-10% fewer. That's a real cost difference. We use the right tokenizer per model automatically.

How much does a typical chat exchange cost?

A 1,000-token prompt with a 200-token reply on GPT-4o: $0.0025 input + $0.002 output = $0.0045 per call, or $4,500 per million calls. Use the calculator above with your actual prompt to get the real number.

Why are some prompts cheaper on Claude than GPT-4o?

Different per-token prices and different tokenizers. Claude Sonnet is $3/$15 per million; GPT-4o is $2.50/$10. For input-heavy workloads with a short reply, GPT-4o usually wins on cost. For balanced or output-heavy workloads, Sonnet often wins. The calculator above shows the exact split for your prompt.

Compare GPT-4o to other models

Detailed comparisons