Free
$0 free
- ~4.4M tokens a month
- ~1.1M tokens a week
Accept turns on our own product analytics and lets the daily.dev ad pixel report to daily.dev; Decline keeps both off. The Reddit ad pixel runs either way, on our public pages and when you sign up or buy, to measure our ads. The daily.dev pixel loads on our public pages and at sign-up too, but reports nothing unless you accept. Ad pixels may set their own cookies and never see your chats. See our Privacy Policy.
Claude, GPT, Gemini, Grok, GLM and more in one place: chat with web sources, deep research, docs and slides, images and video, voice, and one API key. Start free, no card needed.
Scroll or use the arrow keys: 16 slides.Scroll, or swipe sideways for the next slide: 16 slides.
Product tour
Pick a star to go there.
35 models in one chat: Claude, GPT, Gemini, Grok, GLM and more.
Change model at any point in a conversation, without starting over.
Long threads and large documents stay in view on models with very large context windows.
Choose how hard a model thinks before it answers, on models that support it.
Switch on Fast for quicker replies from the models that offer it.
Guest chat with GPT-6 Luna works without an account.
Frontier models
Most models are trained for a consumer audience, so they decline work that is perfectly legitimate. A de-risked model is a modified version of the original: its weights have been edited to take out that trained reflex to say no, so it refuses far less.
For adults (18+). Use is governed by our Terms of Service, which prohibit illegal and harmful use.
De-risked models
De-risked DeepSeek
Fewer refusals (18+): follows your instructions more closely. Vani's usage rules still apply.
De-risked Z.AI
Fewer refusals (18+): follows your instructions more closely. Vani's usage rules still apply.
De-risked Moonshot AI
Fewer refusals than standard Kimi K3.1 Pro, more conservative than GLM 5.3 Derisked.
Vani searches the web first and cites the sources it used.
Attach PDFs, Word documents, CSV files and images, then ask about them.
Replies can draw charts and diagrams, typeset formulas and offer quick choices you can click.
Fork the conversation at any message to try another direction; the original stays intact.
A conversation that is not added to your history, for quick questions.
Hover or tap a message to see when it was sent, and copy, edit, retry or fork it.
A one-time note in the chat when something new ships, linked to the full post.
Chat with sources

A slower, deeper search that reads more sources and keeps each one beside the answer.

Vani writes documents, slide decks and one-page designs you can edit, present, export and share with a link.



Camera moves, effects, transitions, looks and photo tools: pick a preset, add a photo or a few words, and Studio writes the prompt. 313 of them are on the Free plan.
Ready-made looks, camera moves and effects for your photos and videos.
Talking Photo: Your photo speaks: type a script and pick a voice, or add your own recording. From the Plus plan.
3 s
3 s
3 s
3 s
3 s
Real results from Vani Studio presets.
Make an image or a video clip from a prompt or a photo, then edit it, animate it or ask for variations.
Generate an image from a prompt, or upload one and describe the change.
Make a clip from text, animate an image, or set its first and last frames.
Every image and video you make is kept in one place.
Turn text into speech, music or sound effects, and an image into a 3D model, in the same chat.
Image models include GPT Image and Grok Imagine; video models include Seedance and Grok Imagine.
Images, video, audio and 3D


Tap the mic to dictate, and Vani writes what you say into the message box.
Tap Listen on a reply to hear it read aloud.
Voice


ListenPause
Illustration
Vani runs Python on your files in a private, isolated sandbox to analyse data and make charts, spreadsheets, documents and slides.
Point Codex, Claude Code, Cursor, Cline or OpenCode at the Vani API, with a setup guide for each.
Guides for Xcode, VS Code, JetBrains, Microsoft Foundry and the official OpenAI and Anthropic SDKs.
Add Vani's MCP server to Claude, ChatGPT, Cursor, Claude Code, Codex or Gemini CLI to create images and video and run research there.
Add source files to a project; search matches names across styles, so computeBackoff finds compute_backoff.
Vani opens and reads web pages for you, and asks before it submits a form.
Publish an app you build in chat as a live web app.
Chart monthly revenue from sales.xlsx and send me the summary.
Ran Python
sales = pd.read_excel("sales.xlsx")
sales.groupby("month").revenue.sum() .plot.bar()revenue-chart.pngsummary.xlsx
Illustration
Keep your work together, and let Vani pick up where you left off.
Keep related chats, instructions and files together; Vani answers from the files and shows which ones it used.
Vani remembers useful details across chats. See, edit or turn it off at any time.
Save instructions you reuse, import your GPTs and Gems, or add skills from the community gallery.
Connect services and your own MCP servers so Vani can read from them and make changes you approve; it suggests one when it would help.
Run a prompt on a schedule you choose; results arrive in your Updates.
A look back at your month: active days, the models you used most and what you made.
Describe a multi-step job; Vani plans it and works through it while you are away, with a live timeline.
Connected services that have their own interface can show it right in the conversation.
Projects, skills and tasks
Vani guides you with questions and hints instead of handing over the answer.

Learning
Illustration
Every plan publishes how many tokens you can use with each model, per month and per week. Pick a model to compare.
Plans and usage
Token figures are estimates from each plan's allowance and the model's rate, rounded down; actual usage depends on your input and output, context length and reasoning. Every model on every plan.
$0 free
$18 / month
$78 / month
$188 / month
$464 / month
Weekly and billing-period progress with reset times, in plain numbers.
Usage shows as a percentage, with weekly and monthly progress and reset times. A week can use up to about 23% of the month, so the month lasts.
Week used up early? On Plus, Pro, Max and Ultra, Burst turns the weekly limit off until your plan renews and makes 80% of what is left of your month available right away. Once a month.
Need more? On a paid plan, add Extra usage with a one-time payment. Before you pay, you see what it adds as a share of a month of your plan. It lasts 12 months and is used after your plan's allowance.
Prepaid, pay as you go: eligible models cost 30% below their verified official rates.
Developer API
Keep your SDK and change the base URL: Chat Completions, Responses and Messages.
Know when a model joins or leaves the API: follow the changelog as an RSS or JSON feed, poll the changes API, or have each change sent to your server as a signed webhook.
Add credit and pay per request. Eligible models cost 30% below their verified official rates.
Generate and edit images and make video clips with the same key.
On models with a cached-input price, prompt tokens served from cache are billed at that lower rate.
Give any model remote MCP tools through the Responses and Messages APIs.
Usage analytics, the cost of every request, request logs, and your limits and tier.
Rate-limit tiers follow your lifetime deposits; each tier raises your limits.
Create API keys billed to your plan in Developer settings (Plus and higher plans).
Publishers can submit an MCP server to the Connectors directory; each one is reviewed before it is listed.
| Model | Input | Output |
|---|---|---|
| Claude Sonnet 5.5 | $1.40official | $7.00official |
| GPT-6.1 Sol | $1.40official | $7.00official |
| Gemini 3.1 Pro | $1.40official | $8.40official |
| Grok 4.7 | $1.40official | $4.20official |
| Kimi K3 | $2.10official | $10.50official |
import os
from openai import OpenAI
client = OpenAI(
base_url="https://api.vani.ai/v1",
api_key=os.environ["VANI_API_KEY"], # a vani_ key
)
reply = client.chat.completions.create(
model="anthropic/claude-sonnet-5.5",
messages=[{"role": "user", "content": "Hello"}],
)
print(reply.choices[0].message.content)curl https://api.vani.ai/v1/chat/completions \
-H "Authorization: Bearer $VANI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"anthropic/claude-sonnet-5.5","messages":[{"role":"user","content":"Hello"}]}'curl https://api.vani.ai/v1/messages \
-H "x-api-key: $VANI_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{"model":"openai/gpt-6.1-sol","max_tokens":256,"thinking":{"type":"disabled"},"messages":[{"role":"user","content":"Hello"}]}'Privacy
We do not use your prompts or completions to train models.
Memory is yours: see it, edit it or turn it off at any time.
Temporary chats are not added to your history, but usage, billing and security records of each request are kept.
Export a copy of your account's data, or delete the account, from Settings.
Our own analytics, and the daily.dev ad pixel's reporting, run only if you accept cookies. The Reddit ad pixel measures our ads either way; ad pixels never see your chats.
Sign up with your email, no card needed, or try guest chat without an account. Upgrade only when you need more.
See plans and pricesBrowse the modelsRead the docsBack to vani.ai
Start free
Guest chat works in your browser with no sign-up.
A coding agent in your editor: it edits files with a diff to review and runs commands only when you approve.
For macOS, Windows and Linux.
Vani from your terminal.
For iOS and Android.