Pricing
Free to try forever. Upgrade when your work is.
For curious explorers.
$0
For scaling validated experiments.
$32/mo
Cancel any time. First month refundable, no questions.
For teams of 3 or more.
$55/seat/mo
Includes a guided demo + 14-day full-Team eval for your org.
For heavy-iteration teams and quants.
$85/seat/mo
Includes a guided demo + 14-day full eval for your org.
Enterprise starts at $18k ACV: SSO, SCIM, custom retention, on-prem licensing. Talk to us →
Honest guidance. We'd rather you pick right than pick expensive.
| Team size | What to buy | Why |
|---|---|---|
| 1 person | Pro · $32/mo | Solo user, full features. |
| 2 people | 2× Pro · $64/mo | Cheaper than Team. Share via public or private links. |
| 3–5 people | Team · from $165/mo | Pooled GPU + shared workspace + one invoice. |
| 6–15 people | Team · $330–825/mo | Pooling value compounds; admin controls matter. |
| Heavy iteration (any size) | Team Plus · from $255/mo | A100/H100, priority queue, larger pool. |
| 16+ people | Enterprise · custom | SSO, compliance, dedicated support. |
A 3-person Team gets 30 pooled GPU hours/month. That's more than 3× Pro (which gives 5 each, 15 total) — because pooling lets your heaviest user borrow from lighter ones. The bigger your team, the more pooling compounds.
Add extra training time for heavy workloads — long runs, complex models, or large datasets. Expires 12 months after purchase.
$10
A few extra hours
Top up when a single run spills over
$25
A full workload
Keep iterating through a bigger project
$60
Sustained training
For long runs, complex models, or large datasets
Every published model gets a live prediction endpoint. Call it from our Python / JavaScript SDK.
ChatGPT gives you code you still have to run, debug, and deploy. MathExec gives you the working model. Done. You compose the formula visually, point at data, and get a trained model + live SDK endpoint in under a minute.
You can. Every data scientist who tried MathExec said the same thing: 'this is what I do in Python, but in 5% of the time.' If you build models once a quarter, Python is fine. If you iterate weekly, the setup cost adds up fast.
You install our SDK (Python + JavaScript) into your actual application and call the model from real code. Anything composed here runs in production. It's a pro tool with a simple interface, just like Figma.
No automatic trial — Free is genuinely free. Sample your data, prototype on CPU, iterate as much as you want. When a working experiment is ready for production data or cloud GPU, upgrade to Pro. Pro is monthly with first-month refundable, so the risk is low.
Yes. Email sales@mathexec.com — we'll set up a guided demo and grant your org 14 days of full Team access for the eval period.
Whenever CPU training feels slow. Two things drive that: data volume and model complexity. A simple formula on 500k rows is slow on CPU. A transformer on 500 rows is also slow on CPU. A simple formula on 500 rows is instant — stay on CPU. We nudge you toward GPU when we see slowness coming (big data, heavy architecture, long runs, or past trainings that took a while). Rule of thumb: if training feels like waiting, GPU is 5–10x faster.
CPU training runs locally in your browser for free. Cloud GPU is optional on Pro+. You can also export standalone PyTorch code and train anywhere you like.
You get a banner warning at 70 and 90 percent before you hit anything. At 100 percent we surface an upgrade link or credit-pack option. Nothing silently breaks.
Yes. Sign up with a verified .edu email and you get 50% off Pro automatically at checkout.
Enterprise customers can. Starts at $18k ACV with SSO (Okta / Azure AD / SAML 2.0), SCIM, custom retention, and an on-prem license path. Talk to us.
Sample your data, prototype on CPU, iterate freely. Upgrade to Pro when you're scaling a working experiment to production.
Use it free →