Guide
Prompt optimization
In llmwise every message has a fixed price, so the cheapest message is the one you don't have to send twice. These habits get a usable answer in fewer tries, on any model.
Eight habits
Say who it's for and why
“A two-line reply to a customer who was double-charged” beats “write a reply”. Audience and purpose shape everything else.
Give the material, not a description
Paste the text or attach the file instead of describing it. The model can't improve what it hasn't seen.
Show an example
One example of what good looks like is worth a paragraph of adjectives.
Ask for a format
A table, three bullets, a JSON object, a numbered plan: ask for the shape you'll use.
Set the length
Word counts and “no more than five items” keep answers usable, and short.
Let it ask first
For a big task, ask the model what it needs to know before it starts. One round of questions saves several rewrites.
Split big jobs
Ask for a plan, check it, then have it work through one step at a time.
Save what you repeat
Instructions you repeat in every chat belong in a persona: a name, instructions and a default model.
Before and after
Before:
Write something about our new pricing.
After:
Write a 120-word email to existing customers announcing our new pricing (attached). Lead with what stays the same, then the one change that affects them. Friendly and plain, no exclamation marks. End with a link placeholder for questions.
The second prompt names the audience, the length, the order, the tone and the material. It's longer to write and much shorter to fix.
In llmwise
Personas hold instructions you'd otherwise retype, with a default model.
Skills hold longer instructions and reference files the model reads when a task needs them.
Everyday models don't touch your monthly allowance on a paid plan. Try a clear prompt there first.
When you're done with a topic, start a new chat: see what a context window is.
Questions
Do longer prompts cost more in llmwise?
No. In llmwise a message has a fixed price, however long it is. What does change the price is a long chat: once the conversation passes 64k tokens, each reply counts as 2 (as 4 past 128k), and the composer tells you before you send.
Does a better prompt beat a better model?
Often enough to try first. A clear prompt on an everyday model regularly does the job; when it doesn't, switch to a stronger model in the same chat and it sees everything you've said.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.
See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.