How pricing works

The short version

• You always see the price before you send. Next to the send button, llmwise shows how many messages you have left on the model you picked. For example: "Claude Opus 5.5 · 62 messages left this month".
• Every message has a fixed price. A reply costs the same whether it's one line or three pages. What you see before sending is exactly what's used.
• Fast models refill every day on paid plans. You get a fresh daily number at 00:00 UTC, up to your plan's fair-use limit.
• Free is a one-time trial: 5 messages on Claude Haiku 5.5. Every other model is on a paid plan.
• Every limit is published on this page. There are no hidden caps and no "5x more usage" without a number.

Plans at a glance

Prices are in US dollars per month. Tax is added at checkout, depending on where you live.

Free is 5 messages to try llmwise, one time. They work on Claude Haiku 5.5; every other model needs a paid plan. There's no daily refill and no monthly allowance.

Plus is half of Pro. It has every model but Claude Fable 5.1 and GPT-6 Astra, half Pro's monthly allowance and fast messages, and the same longest reply. Research reports, code runs, backend apps, connectors and MCP servers are on Pro and higher plans.

The premium rows share one allowance. The numbers above are how many messages you'd get if you used only that model. Mix freely: using a bigger model simply uses your allowance faster. The next section explains exactly how.

How your monthly allowance works

Each paid plan comes with one monthly allowance for premium models:

Every premium model has a size, meaning how much of the allowance one message uses. The size reflects what the model costs to run (see "How we set these numbers" below).

You never have to do this math. llmwise does it for you and shows the result as plain messages:

Example: Pro, start of the month (allowance 250)
The model picker shows: Fable 31 left · Opus 62 left · Sonnet 125 left · Haiku 250 left.

You send one message to Claude Fable (size 8). Your allowance goes from 250 to 242, so every count updates together:
Fable 30 left · Opus 60 left · Sonnet 121 left · Haiku 242 left.

The count next to the send button always shows the model you have selected. Switch models and it updates instantly.

Your allowance renews on the same day each month as your subscription (your billing date), not on the 1st.

Fast models: a daily number that refills

GPT-6 Luna, Claude Haiku 5.5, DeepSeek V4.1 Flash and GLM 5.3 Flash are our fast models. They're quick, capable and cheap to run, so they don't touch your monthly allowance. Instead every paid plan gets a daily number:

The daily number refills at 00:00 UTC every day. Unused fast messages don't carry over to the next day, but you get a fresh set every day: use them all and you're out only until the day ends, up to the fair-use limit below.

What counts as one message

• One message = one reply from the model. The price is fixed per message and shown before you send, however long the reply is (up to your plan's longest reply).
• Regenerating a reply, or editing your message and sending it again, is a new message.
• Stopping a reply partway still counts, because the model has already done the work.
• Web search: when the model searches the web, or reads a link you give it, each search counts as one more message of the model you're using. On Free, web search works inside your 5 trial messages. You control searching with the Search button next to the message box:
  • Auto (default): the model decides when a search helps.
  • Always: it searches before every answer.
  • Never: no web access at all, so no search charges.
• Tool steps are included. When the model uses a connector (Gmail, Notion and others) or one of your MCP servers, those steps are part of the message price. Actions that send or change something always ask you first.

Long chats

Every time you send a message, the model re-reads the whole conversation so far. A very long chat therefore costs much more to answer than a short one. To keep prices fixed and fair, very long chats count a little more, and you always see this before you send:

Tip: starting a new chat for a new topic keeps every message at its normal price. The count next to the send button shows when a chat has grown long.

When a chat gets long, you can continue in a new chat and your context comes along as a summary, with each message back at its normal price.

Tools and what they cost

Some tools do much more work than a normal reply. Each has a fixed price from your monthly allowance, shown before it runs. Anything that takes a while asks for your approval first.

Research reports, code runs, backend apps, connectors and MCP servers are available on Pro and higher plans. Images are made with Create, on every plan.

Images in chat. An image you ask for in chat is made with Create, as in the studio: it's paid in credits (see "Creating images, video, voice and music" below), and it keeps going if you leave the page. Press Create and choose Image to pick the model and see its price before you send. Models made through kie.ai send kie.ai only that image's prompt, never your chat; the Privacy Policy says what kie.ai may do with it.

Creating images, video, voice and music

Create (images, video, voice, music and upscaling) is paid in credits.

• 1 credit is the price of one Claude Haiku 4.5 message from your monthly allowance.
• 1 credit is about 5¢ (the credit pack price: $10 for 200 credits). Prices are always in credits; the money is a guide.
• You see the price before you press Create. What you see is what's used.
• Failed, blank, refused or blocked costs nothing. If nothing is made, nothing is used.
• Free gets 10 credits, once per email address. There's no refill. They make images and video with the open models: Z-Image, FLUX.2 Klein 4B, Qwen-Image and Wan 2.2 Turbo. They don't pay for any other model, nor for voice, music, talking photos, 3D, upscaling, video edits, dubbing or reels: those need a credit pack or a paid plan.
• On Free, a credit pack adds 200 credits for $10 (plus tax). It lasts 12 months and pays for creating only, on any model, once your free credits can't pay, oldest pack first. It's priced in US dollars, for visitors from India too; an account that has had a plan billed in rupees can't buy one.
• Paid plans use the monthly allowance first, then top-ups. The same allowance covers your messages and Create.

Every prompt and photo is checked before anything is made, and every image and video is checked again before you see it. The Terms (section 9) set out the rules.

Top-ups

If you run low before your allowance renews, you can buy a top-up on any paid plan billed in US dollars (on Free, the same $10 buys a credit pack for creating, above):

$10 (plus tax) adds 200 to your allowance. That's 25 Fable or GPT-6 Astra messages, 50 Opus, 100 Sonnet, GPT-6 Sol, Gemini Pro or Kimi, or 200 Haiku-class messages, or any mix.

• Top-ups are used after your plan's own monthly allowance.
• They last 12 months from purchase.
• If you cancel, you can keep using your top-ups for 30 days after your plan ends, for messages only (fast models count 1; web search needs a plan), within your last billing period's fair-use limit. Anything left then waits until 12 months after purchase, in case you come back.
• Top-ups are a one-time purchase, never a subscription.

Limits we publish, so you never hit a surprise

If we ever need to change a limit, we update this page first and say so in the app. A message's price never changes after you've seen it.

Fair-use limit

Every paid plan includes up to a fixed amount of AI cost each billing period, measured at the list prices in "How we set these numbers": Plus $3.75, Pro $7.50, Max $20, Ultra $42, Studio $85. Using every message on your plan at typical sizes stays under it; very large messages and heavy research use it faster.

We tell you when you've used 80%. If you reach it, every message pauses, the fast models' too, until your allowance renews or you buy a top-up, which adds $4 of room. We never pause you without saying why.

After you cancel, your top-ups' 30 days count toward your last billing period's fair-use limit. If you reach it then, they pause for good: nothing renews, and a top-up bought without a plan pays for creating only. Anything left waits until they expire, in case you come back.

How we set these numbers

We pay the model makers per token: a token is roughly ¾ of a word. The model reads your conversation (input) and writes a reply (output). Output costs more than input. Here are the public list prices we pay, per million tokens, and what a typical message costs us. A typical message is about 4,000 tokens read and 700 written.

The rule: one unit of your allowance covers about $0.01 of a typical message. A model's size is its typical cost rounded up to the next of 1, 2, 4 or 8 units. That's why Claude Fable, which costs about ten times more to run than Claude Haiku, uses 8 where Haiku uses 1.

* Kimi K3's typical cost would round up to size 4. We price it at 2 on purpose, to make it easy to try. Grok 4.7's would round up to size 2 since its price rose to $2 / $6; we keep it at 1 on purpose, so your Grok count didn't change.

Mistral Large 4's prices are a sale price, half its $1.36 / $4.18 list price. At the list price a typical message costs $0.0084: still size 1.

Gemini 4 Argon's prices are Google's standard price from December 1, 2026. Until then Google charges an introductory $2 / $10 ($0.0150 a typical message). We price it at its standard size now, so your Gemini 4 Argon count doesn't drop on December 1.

Claude Haiku 5.5's prices are for prompts up to 100,000 tokens. Anthropic charges $0.50 / $2.50 for a prompt over that, and that's the cost a long chat's message on it counts toward the fair-use limit. A message's price doesn't change: a fast message counts 1, or 2 or 4 in a long chat, like every other.

Why a price per message instead of per token? Per-token billing means you only find out the cost after the reply. We'd rather you know before you send. Some messages cost us more than their price and some less; we take that risk, not you.

Prices can change. Model makers sometimes change their prices. If a change moves a model to a different size, we'll update this page and tell you in the app before it takes effect. A message you've already been shown a price for is never re-priced.

Upgrading, downgrading and cancelling

• Upgrade anytime. You're charged the difference for the rest of your billing month, plus tax. We show the exact amount before you confirm, whenever we can calculate it. Your new allowance starts right away.
• Downgrade anytime. The change happens at the end of your billing month. You keep your current plan until then.
• Cancel anytime in Settings → Billing, in two clicks. You keep your plan until the end of the month you've paid for, and nothing in your account is deleted. We'll ask why you're leaving, but you don't have to answer.
• Money back in the first 7 days. First month not working for you? Write to hello@llmwise.ai within 7 days of your first charge and we refund it. No questions.
• Your chats, files, memory and projects stay yours on every plan, including Free.

Where to see your usage

• Next to the send button: messages left on the model you picked.
• In the model picker: messages left on every model.
• Settings → Billing: your plan, allowance left, top-ups, and your renewal date.