AI Token Counter · Multi-Model

Paste any prompt or drop a file — get exact GPT tokens, approximate Claude/Gemini counts, and USD pricing for 14 frontier models including GPT-5, o3, o4-mini, Gemini 2.5, and Claude Opus 4.7.

Pricing snapshot: 2026-04-24 (USD per 1M tokens, public list prices).

Input

241 chars · 311 bytes

Estimated output tokens|

Pricing mode

Prompt caching (~90% off cached input)Claude · GPT

Batch API (50% off)OpenAI · Anthropic

Token counts & cost per model

Cheapest: Gemini 1.5 Flash

NEW

GPT-5

OpenAI

o200k_base · exact

60tokens

Input ($10/1M): $0.00060
Output ($40/1M): $0.0200
Total: $0.0206

GPT-4o

OpenAI

o200k_base · exact

60tokens

Input ($2.5/1M): $0.00015
Output ($10/1M): $0.00500
Total: $0.00515

GPT-4o mini

OpenAI

o200k_base · exact

60tokens

Input ($0.15/1M): $9.00e-6
Output ($0.6/1M): $0.00030
Total: $0.00031

NEW

o3

OpenAI

o200k_base · exact

60tokens

Input ($10/1M): $0.00060
Output ($40/1M): $0.0200
Total: $0.0206

NEW

o4-mini

OpenAI

o200k_base · exact

60tokens

Input ($1.1/1M): $6.60e-5
Output ($4.4/1M): $0.00220
Total: $0.00227

GPT-3.5 Turbo

OpenAI

cl100k_base · exact

80tokens

Input ($0.5/1M): $4.00e-5
Output ($1.5/1M): $0.00075
Total: $0.00079

Claude Opus 4.7

Anthropic

approx (Claude) · approx.

78tokens

Input ($15/1M): $0.00117
Output ($75/1M): $0.0375
Total: $0.0387

Claude Sonnet 4.6

Anthropic

approx (Claude) · approx.

78tokens

Input ($3/1M): $0.00023
Output ($15/1M): $0.00750
Total: $0.00773

NEW

Claude Sonnet 3.7

Anthropic

approx (Claude) · approx.

78tokens

Input ($3/1M): $0.00023
Output ($15/1M): $0.00750
Total: $0.00773

Claude Haiku 4.5

Anthropic

approx (Claude) · approx.

78tokens

Input ($0.8/1M): $6.24e-5
Output ($4/1M): $0.00200
Total: $0.00206

NEW

Gemini 2.5 Pro

Google

approx (Gemini) · approx.

85tokens

Input ($1.25/1M): $0.00011
Output ($10/1M): $0.00500
Total: $0.00511

NEW

Gemini 2.5 Flash

Google

approx (Gemini) · approx.

85tokens

Input ($0.15/1M): $1.27e-5
Output ($0.6/1M): $0.00030
Total: $0.00031

Gemini 1.5 Pro

Google

approx (Gemini) · approx.

85tokens

Input ($1.25/1M): $0.00011
Output ($5/1M): $0.00250
Total: $0.00261

LOWEST COST

Gemini 1.5 Flash

Google

approx (Gemini) · approx.

85tokens

Input ($0.075/1M): $6.37e-6
Output ($0.3/1M): $0.00015
Total: $0.00016

Token counts for Claude, Gemini, and Llama are in-browser approximations; use each vendor’s official tokenizer or count_tokens API for exact billing. Pricing as of 2026-04; subject to change.

Korean token efficiency · same sentence, different models

Same Korean paragraph tokenized by each model. Lower tokens/char = cheaper for Korean.

Model	Tokens (KO)	Tokens (EN)	KO tokens/char	Note
GPT-4o (o200k_base)	49	34	0.52	Exact · best for Korean among GPTs
GPT-4 / 3.5 (cl100k_base)	92	34	0.98	Exact · older BPE, less Korean-optimized
Claude 4.x	79	34	0.84	Approx · similar to cl100k for Hangul
Gemini 1.5	59	53	0.63	Approx · highly efficient on Korean per Google docs
Llama 3.1	124	34	1.32	Approx · byte-level Hangul, worst ratio

Cost-saving tips

Reuse the same system message across requests to unlock prompt caching (~90% discount on cached prefix for Claude and GPT).
Use Batch API for non-realtime workloads — 50% off on OpenAI and Anthropic.
Set a strict max_tokens to cap runaway outputs — output tokens are 4–5x more expensive than input.
For Korean-heavy pipelines, Gemini 2.5 Flash and GPT-4o mini give the lowest $/KO-char today.
Strip markdown tables, emoji runs, and repeated whitespace before sending — they bloat BPE tokens.
Route low-stakes traffic to Haiku / Flash / mini; reserve Opus and GPT-5 only for reasoning-heavy turns.

Learn how tokenizers work in the guide, or jump to the FAQ.