Your key, your bill
You connect your own OpenAI, Anthropic, Google, xAI, DeepSeek, Qwen or MiniMax key. Usage is billed by that provider at their price - RolChat adds nothing on top.
What does AI support actually cost?
Most platforms bill you a fixed price for every conversation their AI resolves. RolChat runs on your own provider key, so you pay the provider directly. Here is the arithmetic, on published token prices.
Calculator
Your numbers, real provider prices
Start from a scenario close to yours, adjust it, and see what each model would cost you - for written answers and for spoken ones.
Start from:
Your scenario, cheapest model$1.60OpenAI · GPT-5 Nano
The same work, billed per resolution$990.00
The difference$988.40
Written answers
OpenAI23$0.0003
| Model | Per answer | Per month |
|---|---|---|
| GPT-5 Nano | $0.0003 | $1.60 |
| GPT-4.1 nano | $0.0005 | $2.60 |
| GPT-6 Luna | $0.0006 | $2.75 |
| GPT-5.6 Luna | $0.0012 | $5.80 |
| GPT-5.4 Nano | $0.0012 | $5.88 |
| GPT-5 Mini | $0.0016 | $8.00 |
| GPT-4.1 mini | $0.0021 | $10.40 |
| GPT-5.4 Mini | $0.0044 | $21.75 |
| o4-mini | $0.0057 | $28.60 |
| GPT-5 | $0.0080 | $40.00 |
| GPT-5.1 | $0.0080 | $40.00 |
| GPT-4.1 | $0.010 | $52.00 |
| GPT-6 Sol | $0.011 | $55.00 |
| GPT-6.1 Sol | $0.011 | $55.00 |
| GPT-5.2 | $0.011 | $56.00 |
| GPT-5.6 Terra | $0.012 | $58.00 |
| GPT-5.4 | $0.014 | $72.50 |
| GPT-5.6 Sol | $0.022 | $110.00 |
| GPT-5.5 | $0.029 | $145.00 |
| GPT-6 Astra | $0.055 | $275.00 |
| GPT-5.2 Pro | $0.13 | $672.00 |
| GPT-5.4 Pro | $0.17 | $870.00 |
| GPT-5.5 Pro | $0.17 | $870.00 |
Anthropic13$0.0055
| Model | Per answer | Per month |
|---|---|---|
| Claude Haiku 4.5 | $0.0055 | $27.50 |
| Claude Sonnet 5 | $0.011 | $55.00 |
| Claude Sonnet 5.5 | $0.011 | $55.00 |
| Claude Sonnet 4.5 | $0.017 | $82.50 |
| Claude Sonnet 4.6 | $0.017 | $82.50 |
| Claude Opus 5.5 | $0.022 | $110.00 |
| Claude Opus 4.5 | $0.028 | $137.50 |
| Claude Opus 4.6 | $0.028 | $137.50 |
| Claude Opus 4.7 | $0.028 | $137.50 |
| Claude Opus 4.8 | $0.028 | $137.50 |
| Claude Opus 5 | $0.028 | $137.50 |
| Claude Fable 5 | $0.055 | $275.00 |
| Claude Fable 5.1 | $0.055 | $275.00 |
Google9$0.0005
| Model | Per answer | Per month |
|---|---|---|
| Gemini 2.5 Flash-Lite | $0.0005 | $2.60 |
| Gemini 3.1 Flash-Lite | $0.0015 | $7.25 |
| Gemini 2.5 Flash | $0.0020 | $9.75 |
| Gemini 3.5 Flash-Lite | $0.0020 | $9.75 |
| Gemini 3.6 Flash | $0.0041 | $20.63 |
| Gemini 3.7 Flash | $0.0041 | $20.63 |
| Gemini 3.8 Flash | $0.0041 | $20.63 |
| Gemini 2.5 Pro | $0.0080 | $40.00 |
| Gemini 3.5 Flash | $0.0087 | $43.50 |
xAI3$0.0010
| Model | Per answer | Per month |
|---|---|---|
| Grok 4.1 Fast | $0.0010 | $4.75 |
| Grok 4.20 | $0.0098 | $49.00 |
| Grok 4 | $0.017 | $82.50 |
DeepSeek2$0.0016
| Model | Per answer | Per month |
|---|---|---|
| DeepSeek Flash (V4.1) | $0.0016 | $7.80 |
| DeepSeek V4 Pro | $0.0065 | $32.34 |
Qwen4$0.0012
| Model | Per answer | Per month |
|---|---|---|
| Qwen3.5 Flash | $0.0012 | $6.02 |
| Qwen3 Plus | $0.0033 | $16.62 |
| Qwen3.5 Omni | $0.0075 | $37.60 |
| Qwen3 Max | $0.017 | $82.50 |
MiniMax7$0.0016
| Model | Per answer | Per month |
|---|---|---|
| MiniMax M2 | $0.0016 | $7.80 |
| MiniMax M2-Her | $0.0016 | $7.80 |
| MiniMax M2.1 | $0.0016 | $7.80 |
| MiniMax M2.5 | $0.0016 | $7.80 |
| MiniMax M2.7 | $0.0016 | $7.80 |
| MiniMax M3 | $0.0016 | $7.80 |
| MiniMax M2.7 highspeed | $0.0031 | $15.60 |
| What the same work costs elsewhere | Per conversation | Per month |
|---|---|---|
| A platform that charges for every conversation its AI resolves | $0.99 | $990.00 |
Spoken answers
Transcription, speech and voice agents, priced per minute of conversation. Speech works in chat today - a voice message transcribed, an answer read aloud - and telephony is coming soon.
OpenAI6$0.0030
| Model | Does | Per minute | Per month |
|---|---|---|---|
| GPT-4o Mini Transcribe | Speech to text | $0.0030 | $1.50 |
| GPT-4o Transcribe | Speech to text | $0.0060 | $3.00 |
| Whisper | Speech to text | $0.0060 | $3.00 |
| GPT-4o Mini TTS | Text to speech | $0.014 | $6.75 |
| TTS-1 | Text to speech | $0.014 | $6.75 |
| TTS-1 HD | Text to speech | $0.027 | $13.50 |
Google2$0.024
| Model | Does | Per minute | Per month |
|---|---|---|---|
| Google Cloud STT v2 | Speech to text | $0.024 | $12.00 |
| Google Cloud TTS (Studio voices) | Text to speech | $0.014 | $7.20 |
ElevenLabs4$0.054
| Model | Does | Per minute | Per month |
|---|---|---|---|
| Eleven Flash v2.5 | Text to speech | $0.054 | $27.00 |
| Eleven Turbo v2.5 | Text to speech | $0.054 | $27.00 |
| Eleven Multilingual v2 | Text to speech | $0.11 | $54.00 |
| Eleven v3 | Text to speech | $0.11 | $54.00 |
Deepgram5$0.075
| Model | Does | Per minute | Per month |
|---|---|---|---|
| Voice Agent API | Voice agent | $0.075 | $37.50 |
| Nova-3 Monolingual | Speech to text | $0.0077 | $3.85 |
| Nova-3 Multilingual | Speech to text | $0.0092 | $4.60 |
| Aura-1 | Text to speech | $0.014 | $6.75 |
| Aura-2 | Text to speech | $0.027 | $13.50 |
AssemblyAI2$0.0025
| Model | Does | Per minute | Per month |
|---|---|---|---|
| Universal-2 | Speech to text | $0.0025 | $1.25 |
| Universal-3.5 Pro | Speech to text | $0.0035 | $1.75 |
Fish Audio3$0.014
| Model | Does | Per minute | Per month |
|---|---|---|---|
| S1 | Text to speech | $0.014 | $6.75 |
| S2 Pro | Text to speech | $0.014 | $6.75 |
| S2.1 Pro | Text to speech | $0.014 | $6.75 |
Text-to-speech is published per 1,000 characters; we convert at 900 characters a minute, roughly a normal speaking pace. A full voice conversation usually needs two of these - transcription plus speech - or a single voice agent that does both.
$0.99 per resolved conversation is a common list price for AI support agents. Output is assumed at 300 tokens per answer.
Provider prices as published on the date shown. Providers change them - and you always pay their current price, never ours. Last checked:
How this works
Why we can show you this at all
You pick the model
Choose a different model for replies, for translation, for summaries. Set a primary and a fallback, and a daily spending cap you cannot exceed.
You see the price before you switch it on
The product shows the same estimate inside the app before you enable an AI feature, so nothing arrives as a surprise on an invoice.
You know where the data goes
Every model in the list carries its provider and jurisdiction. When a customer asks where their conversations are processed, you have an answer.
What moves an AI support bill
Two things decide what AI support costs you: how many answers you generate, and how much context each one carries. Volume is easy to estimate. Context is where teams guess wrong, because the conversation so far, a few knowledge base passages and your own instructions add up to thousands of tokens before the model writes a word.
So the calculator asks for both. At 3,000 tokens of context on a small model, a written answer lands in fractions of a cent. Push context to 12,000 and switch to a flagship, and the same answer can cost twenty times more. That is the whole reason the product lets you pick a different model per task.
The comparison column shows the other pricing model on the market: a flat fee for every conversation an AI resolves, commonly listed at $0.99. A thousand resolved conversations a month is $990 that way, against a few dollars of tokens. The fee is a margin on top of the tokens, and you are paying for both.
Read more
Voice is priced by the minute instead of by the token. Transcription, speech and voice agents each publish their own rate, and a full spoken conversation usually needs two of the three. Speech works in chat today, a voice message transcribed or an answer read aloud; telephony is coming soon.
None of this depends on a RolChat plan. You connect your own provider key, that provider bills you at their price, and the same estimate appears inside the product before you switch a feature on. We never resell tokens, so there is no markup to hide.
Questions about AI cost
Why is this so much cheaper than per-resolution pricing?
Because a resolved conversation is a handful of model calls, and model calls are priced in fractions of a cent. A fixed price per resolution is a margin on top of that, not the cost of it.
What is a realistic context size?
For support with a knowledge base, 3,000-6,000 tokens per answer is typical: the conversation so far, a few knowledge base passages and your instructions. Long documents push it higher.
Do cheaper models handle support well?
For routine questions with a good knowledge base, small models do well and cost a fraction of a flagship. Keep a stronger model as the fallback for the hard cases - that is exactly what the primary and fallback setting is for.
What if I do not have a provider account?
You create one with the provider you prefer and paste the key into RolChat. It takes a few minutes, and the account stays yours - including the usage history and the spending limits.
Are these prices guaranteed?
No. They are the providers' published prices as of the date shown, and providers change them. That is another reason we do not resell tokens: you always pay the current price, not ours.
Try it on your own numbers
Free for 30 days. Connect your provider key and see the real cost from day one.