Home / Tools / Tokenizer comparison

Tokenizer comparison. GPT vs Claude vs Gemini

Paste once. Compare token counts across OpenAI Exact encodings and Approx families for Claude, Gemini, and other hosts. See deltas vs o200k and a rough cost on a representative model.

28 words · 182 characters

TokenizerTokensvs o200kAccuracyExample cost
OpenAI o200k_base
GPT-4o / GPT-4.1 / GPT-5 class Exact
35baselineExact$0.002118
GPT-4.1
Open
OpenAI cl100k_base
Legacy GPT-3.5 / early GPT-4 Exact
35+0.0%Approx$0.000402
GPT-3.5 Turbo
Open
Anthropic (Claude)
Browser Approx. Confirm with count_tokens
46+31.4%Approx$0.002652
Claude Sonnet 5
Open
Google Gemini
Browser Approx. Confirm with countTokens
46+31.4%Approx$0.000654
Gemini 2.5 Flash
Open
Generic Approx
Heuristic for other hosts
46+31.4%Approx$0.000294
DeepSeek V3
Open

Baseline is Exact o200k_base. Approx rows are planning aids, not invoices. Confirm Claude and Gemini with provider count APIs before production budgets.

When the gap matters

English prose often stays within a few percent. Code, emoji, and CJK scripts can diverge much more. Try the sample buttons above, then paste your real prompt.

Why GPT, Claude, Gemini counts differ · Stop using tiktoken for Claude · o200k vs cl100k · Tokens per word · Cost calculator

Related tools

Token visualizer · OpenAI tokenizer · Cost calculator

Related guides

Tokenizer comparison: GPT vs Claude vs Gemini · Why counts differ · o200k vs cl100k · tiktoken and Claude · Tokens per word

Sources and references

Official documentation used for definitions, counting methods, or rate cards. Always confirm critical budgets on the provider page.

FAQ

Why do token counts differ across rows?
Each family uses a different vocabulary and merge rules. Same string in, different integers out.
Which rows are Exact?
OpenAI o200k_base and cl100k_base run Exact in this browser. Anthropic, Gemini, and generic rows are Approx.
Can I use tiktoken counts for Claude?
No for billing or hard fit gates. Confirm Claude with Anthropic count_tokens against the model you call.
What does the cost column use?
A representative catalog model for that tokenizer family, plus your reserved output tokens. Open the cost calculator for your real SKU.