Your files stay on your device

# Context-window fit checker

Will your prompt fit, and how much room is left for the answer? Exact for OpenAI models, a labeled estimate for everything else.
exact counts need the AI worker unavailable here; estimate mode still works · registry snapshot 2026-08-26Check fitClearListed OpenAI model (exact)Other model (estimate)Model

o200k_base encoding · max output 128,000 tokens

Paste a prompt, then press Check fit.
System messageUser message(s), one per lineAssistant message(s), one per lineCompare against (up to 4)gpt-5.6-solgpt-5.6-terragpt-5.6-lunagpt-5.6-cybergpt-5.5gpt-5.5-progpt-5.4gpt-5.4-minigpt-5.4-nanogpt-5.4-progpt-5.3-codexgpt-5.2Frame and compare

- OpenAI rows are exact: the count comes from the model's own encoding, cross-checked against the installed tokenizer package's model catalog. Windows come from that same catalog (snapshot 2026-08-26 ).
- Claude and Gemini have no offline tokenizer here: their rows are a characters÷4 heuristic, labeled _estimate_, and this tool never calls a vendor API to refine them. Use the custom mode with the vendor's published window.
- Providers count differently: some bill the system prompt, tools, and message framing separately, and chat wrappers add tokens per message. Treat every number as a planning check, not an invoice.

- gpt-5.6-sol1,050,000 in
- gpt-5.6-terra1,050,000 in
- gpt-5.6-luna1,050,000 in
- gpt-5.6-cyber400,000 in
- gpt-5.51,050,000 in
- gpt-5.5-pro1,050,000 in
- gpt-5.41,050,000 in
- gpt-5.4-mini400,000 in
- gpt-5.4-nano400,000 in
- gpt-5.4-pro1,050,000 in
- … 33 models in the dated snapshot
Paste, drop, or type to begin. Everything stays on this device.

## More AI tools

- [Tokenizer playground](/ai/tokenizer)
- [API cost estimator](/ai/cost-estimator)
[All AI tools](/ai)

---

Canonical HTML: https://nutter.tools/ai/context-fit/
Markdown version: https://nutter.tools/ai/context-fit/index.md
Plain-text version: https://nutter.tools/ai/context-fit/index.txt
Agent index: https://nutter.tools/llms.txt