Skip to content
RendGet started

THE REND MODEL

rend-1

Writes what you asked for. No compliance desk.

Rend’s own model for content. It writes hooks, scripts and ads the way you briefed them, reads your whole brand library in one request, and looks things up when the facts matter.

tokens it can read at once
1M
tokens in one reply
131K
in, text out
Text + images
per 1M in / out, every plan
$0.50 / $2.00
per 1M it has read before
$0.0125

WHAT IT DOES

The briefs other models water down.

Most AI models are built to be careful. Give one a hard-selling ad and it softens your claim, adds a disclaimer or turns the job down.

rend-1 writes it. Your line stays word for word, your price stays as you wrote it, and the angle stays as sharp as you made it. You’re the editor: what runs, and whether it can run, is your call.

YouWrite a 15-second ad. Keep my line word for word: “The only serum you’ll ever need.”

A general chat model

Rend

80% fewer refusalsthan general chat models on creative and ad briefs

Writes with the best of them.Script writing score out of 100.
  • GPT-6.1 Sol82.4Its call
  • Rend Model · Max81.6Your call
  • Claude Opus 4.880.7Its call
ToneBench writing benchmark, Nov 2026, each model at max effort. Rend Model scored as DeepSeek V4.1 Flash class.

Zero editorial inputdirection and compliance stay with you

INSIDE

Room for the whole brief.

It can read a whole library, reread it for almost nothing, and check the facts before it writes.

  • CONTEXT

    1,000,000 tokens in one request

    A token is about three-quarters of a word, so that’s roughly 750,000 words: the brand book, every script you’ve shipped, the transcripts and the research, all at once. A reply can run to 131,072 tokens.

  • REASONING

    Thinks as hard as you tell it to

    Four levels: none, low, high and max. Use none for a quick caption and max for a launch plan. Thinking uses tokens, so the choice is yours on every request.

    1. none
    2. low
    3. high
    4. max
  • CACHING

    Rereading is nearly free

    When a request starts with material rend-1 has read before, like the same brand library, that part costs $0.0125 per million tokens. That’s at least 98% less than new text, and it happens on its own.

  • LIVE SEARCH In chat and the API

    Checks the web and X before it writes

    If a brief depends on what’s true today, such as a price, a trend or what people are posting, rend-1 looks it up first. Every page and post it read comes back with a link.

  • IMAGES

    Looks at up to 10 images

    Product shots, screenshots, a rival’s ad, a storyboard frame. Each one counts as about 1,000 tokens of reading at most.

  • TOOLS AND JSON

    Fits the code you already have

    It uses the same chat format as OpenAI’s API, so existing clients work. It can call your functions (up to 128 tools), return JSON that matches your schema, and stream the reply as it writes.

X AND META

From what people are posting to an ad that’s running.

rend-1 reads X for the angle and writes the ad. In Studio it can take that ad the rest of the way to Meta.

  1. XWhat people are saying this week
  2. rend-1Writes the hook, script and copy
  3. MetaChecked against the rules, then live

X SEARCH In chat and the API

It reads X before it writes

Ask for hooks on a topic and rend-1 searches the last seven days of posts about it. It gets the posts themselves, with their likes, reposts, replies and views, then writes from what’s landing right now.

  • Narrow it to the accounts you name, or to exact dates
  • 0.5¢ for each post it reads, up to 40 a search
  • Every post comes back with its link

META ADS

It takes the ad to Meta

In Studio, rend-1 writes what a Meta ad needs: the hook and script for the video, the headline, the primary text and the call to action. Rend makes the creative from it.

  • A check against Meta’s current ad rules, whenever you ask for one
  • The check reads your landing page and the ads you already run
  • Boost puts it live on Facebook, Instagram and Threads (early access)
See Rend Boost

HOW IT WORKS

One request, start to finish.

  1. 1You send a brief

    Words, up to 10 images, and your own instructions if you have them.

  2. 2It thinks

    For as long as you allow, from not at all to as hard as it can.

  3. 3It looks things up

    Only when the brief needs today’s facts. It searches the web and X, then reads what it found.

  4. 4It writes

    You get the copy, a link for every source and the exact cost.

  • Open now

    In chat

    Open rend.ai/chat and type. Your chats are saved, and every reply shows its sources and what it cost.

    Open chat
  • Open now

    In Studio

    Choose the Rend model in chat. It runs the whole crew, from research and script to the edit, the sound and the review. Each message shows what it cost.

    Open Studio
  • Early access

    From your own agent

    Already working in Claude or ChatGPT? Connect it to Rend and it can hand the writing to rend-1 in one call.

    Connect an agent
  • Early access

    Through the API

    For your own code. Point an OpenAI client at Rend, set the model to rend-1, and it works.

    See the API

PRICING

Priced by the token. Shown on every reply.

You pay for what it reads and what it writes, from your Rend credit. There’s no minimum and no seat fee, and a request that fails on our side costs nothing.

rend-1

Every plan. Uncensored, the default
Inputwhat it reads
$0.50
Cached inputwhat it has read before
$0.0125
Outputwhat it writes, thinking included
$2.00
USD per 1M tokens, about 750,000 words

rend-1-regular

Every plan. The regular model
Inputwhat it reads
$0.25
Cached inputwhat it has read before
$0.005
Outputwhat it writes, thinking included
$0.80
USD per 1M tokens, about 750,000 words

Search and images

Only when used
  • Web search0.1¢ a search0.1¢ a page read in full
  • X search0.5¢ a post0.5¢ to 20¢ a search
  • Image inputAbout 1,000 tokensat most, at the input price
  • Failed on our sideFree

What will a request cost?

On any plan

$0.0032

About 317 of these for $1.

Thinking counts as output. A search adds its fee, plus the reading of what it found.

API · EARLY ACCESS

Change two lines. You’re on rend-1.

It speaks OpenAI’s chat format at https://rend.ai/api/v1, so the client you already use works. Every reply tells you what it cost.

curl https://rend.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $REND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "rend-1",
    "messages": [
      {"role": "user", "content": "Write three hooks for my ad."}
    ]
  }'
The reply, with its cost
{
  "model": "rend-1",
  "choices": [
    { "message": { "role": "assistant", "content": "1. …" } }
  ],
  "usage": {
    "prompt_tokens": 412,
    "completion_tokens": 388,
    "cost": 0.000982,
    "cost_details": {
      "input": 0.000206, "cached_input": 0,
      "output": 0.000776, "search": 0
    }
  },
  "search_results": []
}
  • Streaming"stream": true sends the reply as it’s written.
  • Search control"search" is "auto", "web" or "off". X search runs only when the prompt asks for posts from X, or when you send "x_posts", which sets how many posts one request reads.
  • Reasoning"reasoning_effort" takes "none", "low", "high", "max". Those are the only levels. Leave it out and you get "high". An OpenAI client that sends "medium" gets "high" too.
  • Tools and JSONIt calls your functions through tools and tool_choice, and response_format gets you JSON, or JSON that fits your schema.
  • Samplingtemperature, top_p, top_k, min_p, seed and the penalty settings are accepted, so existing code runs unchanged. While rend-1 thinks, which is the default, temperature and the penalties change nothing. stop cuts the thinking too, so use it with "reasoning_effort": "none".
  • Your pricesGET /models returns what your account pays. Every request also shows up in Usage.
Request API access

We’re letting accounts in a few at a time. If yours is in, create a key in Settings → Connections.

Questions.

What is rend-1?

It’s Rend’s own model, made for content: scripts, hooks, ads, captions, posts, emails, landing pages and titles, plus the plan for getting them seen. It runs the Rend agent in Studio, and your own code and agents can use it too.

How is it different from using Claude or ChatGPT in Rend?

Studio works with your own Claude or ChatGPT subscription, and it works with rend-1, paid from your Rend credit. The difference is in the writing. rend-1 writes the brief as you gave it. It keeps your lines and doesn’t add disclaimers.

What is a token?

The unit a model counts text in. One token is about three-quarters of a word, so 1,000 tokens is roughly 750 words. You pay for the tokens rend-1 reads, called input, and the tokens it writes, called output.

What is rend-1-regular?

The same model with its usual limits on what it will write, at a lower price. Set the model to "rend-1-regular", or keep "rend-1" and send "uncensored": false. Everything else about a request is the same, and GET /models lists both with their prices.

How does caching work?

On its own. When a request starts with material rend-1 has read before, like the same instructions or brand library, that part costs $0.0125 per million tokens. There’s nothing to turn on and no fee for storing it. Each reply says how many tokens were cached.

What does a search cost?

A web search is 0.1¢, and reading a page in full is 0.1¢. An X search is 0.5¢ for each post it brings back, with a 0.5¢ minimum and a 20¢ maximum per search. What it reads from the results counts as input. To turn search off, send "search": "off".

Is a failed request charged?

Not if the failure was ours. If you cancel a request, or it runs out of credit or time, you pay for the part that ran.

Are there rate limits?

Rend sets none on your account: no cap on requests or on how many run at once. Each request sets aside the credit it could use, so your balance is your limit. The model’s capacity is shared, though, so at a very busy moment a request can come back as a 503. That request costs nothing. Send it again.

Who is responsible for what it writes?

You are. rend-1 follows your direction and keeps your wording, which means the editorial calls are yours, and so is staying within the law, platform rules and ad policies.

Can it make video, music or images?

On its own it reads text and images and writes text. Inside Studio it directs the tools that make video, voice, music and images, so a single brief can still end as a finished cut.

Give it the brief.