AI Model Context & Character Limits
Every model accepts a prompt only up to a certain size. Send more than its context window and it either truncates your prompt or refuses. This table shows the real limits and per-token prices for the models people ask about most, taken directly from the 191 models in the TPEE app (last updated 2026-08-25 09:30 +10:00).
By Jacques Botte, founder of Toptronic®. Last updated 12 September 2026.
Context windows and prices (headline models)
| Model | Context window | Input $/1M | Output $/1M | Released |
|---|---|---|---|---|
| Kimi K3 | 1M | $3 | $15 | 07-16-2026 |
| Claude Fable 5 | 1M | $10 | $50 | 06-09-2026 |
| GPT-5 | 400K | $1.25 | $10 | 08-07-2025 |
| Gemini 3.5 Flash | 1M | $1.5 | $9 | 05-19-2026 |
| DeepSeek V4 Pro | 1M | $0.435 | $0.87 | 04-24-2026 |
| Qwen 3.7 Max | 1M | $2.5 | $7.5 | 06-01-2026 |
| Mistral Large 3 | 256K | $0 | $0 | 05-20-2026 |
| Grok 4.6 | 500K | $2 | $6 | 08-12-2026 |
| GLM-5 | 202K | $1 | $3.2 | 02-12-2026 |
| Llama 4 Scout | 10M | $0.08 | $0.3 | 04-05-2025 |
Prices and windows change often — this table is a snapshot. The full, up-to-date list of all 191 models across 20 providers ships inside the TPEE app, where the API Character Limits panel checks your prompt against the exact limit of the model you've chosen before you send it.
Why the limit matters
A prompt that's too long for the model gets cut off mid-thought — the model never sees the end of your instructions and answers as if they weren't there. A prompt that's just under the limit, meanwhile, may leave no room for the model's own answer. TPEE's character counters and per-model limits let you size the prompt to the model you're actually paying for, so you don't discover the problem after you've spent the tokens.
Get TPEE or read the prompt engineering guide.