StacksFree Verdict: 8/10. Modal’s Starter plan gives you real, recurring monthly serverless GPU and CPU compute with $30 in free credits [1], sub-second cold starts, and zero-YAML config-as-code, with the main catches being 1-day log retention and a 10-GPU concurrency cap. If you’re searching for how to get started with Modal for beginners, the walkthrough below is verified against Modal’s official docs and pricing page.
How This Guide Was Built
This guide is based on Modal’s official documentation and Modal’s official pricing page, both verified on September 4, 2026. We verified the recurring free credits, Starter plan limits, signup flow, first-project steps, and per-second GPU and CPU rates. We did not run Modal hands-on. Last verified: September 2026.
How do I get started with Modal?
You get started with Modal in three steps: create a free account at modal.com, install the Python package with pip install modal, and authenticate with modal setup. After that, you run your first app with modal run file.py. Python is the primary language, but JavaScript, TypeScript, and Go can call Modal Functions.
What do you get on Modal’s free Starter plan?
Modal’s Starter plan costs $0 per month and includes $30 in free compute credits every month [1], which recur as long as your account stays active. You also get 3 seats, 100 containers, 10 GPU concurrency, 200 deployed apps, 5 deployed cron jobs, and 3-version rollbacks, all confirmed on Modal’s pricing page.
How much compute does $30 of free credit buy?
Your $30 monthly credit buys roughly 50 hours on a T4 GPU at $0.59 per hour, or about 544 hours on a 1-core, 1-GiB CPU container at roughly $0.055 per hour. Modal charges per second, so you pay nothing while your code is idle, and containers autoscale to zero when not in use.
How to write and run your first Modal app
You define Modal apps in Python using decorators, with no YAML configuration required. Here is a minimal example that runs a Hugging Face model on a T4 GPU:
import modal
app = modal.App("hello-gpu")
image = modal.Image.debian_slim().pip_install("transformers", "torch")
@app.function(image=image, gpu="t4")
def transcribe(audio_url: str) -> str:
from transformers import pipeline
pipe = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
return pipe(audio_url)["text"]Save that as app.py and run it with modal run app.py. For web endpoints, add @modal.fastapi_endpoint and use modal serve app.py for local development, per the Modal docs.
Modal Starter plan limits: what’s free vs. what costs money?
On the Starter plan, your $30 monthly credit covers compute only; you pay nothing upfront and nothing while idle. Paid features kick in when you exceed Starter limits or need Team features like unlimited seats and 5,000 containers, which cost $250 per month plus compute. Volumes cost $0.09 per GiB per month after a free 1 TiB allocation.
Common mistakes beginners make with Modal
The most common mistake is ignoring the 1-day log retention on the Starter plan, so you need external logging for debugging. Another is forgetting to specify gpu="t4" in your decorator and accidentally running CPU-only containers. Finally, avoid leaving non-preemptible GPU jobs running, since they cost 3× the base rate.
FAQ
Is Modal’s free tier really free?
Yes, Modal’s Starter plan is free with $30 in recurring monthly credits, and you are not charged unless you exceed that credit amount. There is no upfront cost, and you pay per second only for compute you actually use, as stated on Modal’s pricing page.
What GPUs are available on Modal?
Modal offers a range of GPUs including T4, L4, A10, L40S, A100 40GB and 80GB, RTX PRO 6000, H100, H200, B200, and B300. You select one by passing gpu="h100" or similar in your function decorator, with prices starting at $0.000164 per second for a T4.
Can I use Modal with JavaScript or Go?
Yes, you can use JavaScript, TypeScript, and Go to call Modal Functions, run Sandboxes, and manage resources, even though Python is the primary language for defining apps. The Modal documentation shows how to invoke Python-defined Functions from these languages.
Where to go next
Once you have your first Modal app running, compare it with other free compute options like the Groq free tier for ultra-fast inference or the Hugging Face Spaces free tier for hosted demos. You can also explore our roundup of the best free AI APIs to see where Modal fits your stack. If you need plain serverless functions without GPUs, our Deno Deploy free tier guide covers a generous always-free option.
Sources
[1] Modal Pricing — https://modal.com/pricing