PromptMeter
Know what your prompt costs before you send it.
Type a prompt, see token counts and input costs across GPT-4o, Claude, Gemini, Llama, and more — instantly, without making a single API call. A shared Rust core runs the tokenizers offline; proprietary models fall back to a counting endpoint only when you need exact numbers. Visual context-window bars warn you at 80% saturation before you hit a model's limit.
What makes it different.
Instant offline tokenization
HuggingFace tokenizers + tiktoken-rs in a shared Rust core. Sub-100ms from keystroke to count — zero network for 13 open models.
Cost estimation across providers
Input cost side-by-side for every model. Prices stay current via live models.dev refresh — no stale hardcoded numbers.
Context-window guardrails
Visual 80% saturation warning per model before you hit the limit. No surprise truncation mid-inference.
Side-by-side model comparison
Compare token count, cost, and context usage across 17 models at once. Pick the right model before the API call.
Simple, honest pricing.
No ads. No tricks.
- Token counting for open models (offline)
- Cost estimation
- Context-window guardrails
- No API key required for standard models
One-time. Yours forever.
- Proprietary model counting (Claude, Gemini) via API fallback
- Saved prompt history
- Output token projection
What we collect. What we don't.
No ad SDK. Ever.
Nothing to sign into.
Anonymous crash reports only.