FreqAI in the cloud — hosted training and live inference
Train FreqAI and reinforcement-learning models in the cloud and run live inference on your bots — inline on your existing runners, with no GPU job to schedule and no ML infrastructure to manage.
- RL algorithms curated
- 6
- PPO · A2C · DQN · QRDQN · TRPO · MaskablePPO
- GPU jobs to schedule
- 0
- training runs inline on your runner
- Plans with FreqAI
- 2
- Pro & Enterprise
FreqAI in the cloud — hosted training and live inference, explained.
FreqAI cloud means hosted model training and live inference for FreqAI, Freqtrade's machine-learning framework. VolatiCloud runs both on managed infrastructure, so you focus on the model and the reward — not on building and operating an ML pipeline.
That includes FreqAI-native reinforcement learning — PPO, A2C, DQN, QRDQN, TRPO, and MaskablePPO — trained inline on the same runners your bots already use. There's no separate GPU job to schedule and no database migration; the model config rides alongside your strategy. No major bot platform offers managed FreqAI-native reinforcement learning.
Training compute is metered honestly — longer training runs use more credits. Once trained, a model can drive a dry-run bot first, so you can watch its live behaviour on simulated orders before it touches real funds.
From idea to a running bot.
FreqAI is powerful and operationally heavy to self-host — so we host it.
Configure the model
Choose a reinforcement-learning algorithm and shape the reward visually or in code.
Train inline
Training runs on your existing runner — no separate GPU job, no extra infrastructure to stand up.
Run live inference
The trained model drives live or dry-run bots, with inference managed by the platform.
Iterate
Backtest, retrain, and compare models against rule-based strategies on the same metrics.
Built for the way you trade.
FreqAI cloud is for traders who want machine learning in their bots without becoming an MLOps engineer.
ML-curious traders
Use FreqAI and reinforcement learning through a guided UI instead of building a training pipeline.
Self-hosting FreqAI users
Move the heavy training and inference workload onto managed infrastructure.
Quant researchers
Experiment with reward design and algorithms with cloud compute behind you.
- Hosted FreqAI training and inference
- FreqAI-native reinforcement learning
- PPO, A2C, DQN, QRDQN, TRPO, MaskablePPO
- Trains inline on existing runners
- No GPU job, no database migration
- Visual reward builder and code mode
Frequently asked questions.
What is FreqAI?
FreqAI is Freqtrade's built-in machine-learning framework. VolatiCloud hosts it so model training and live inference run on managed infrastructure instead of a stack you build and operate yourself.
Does training need a GPU job?
No. Training runs inline on your existing runner — there's no separate GPU job to schedule and no database migration. The model config travels with your strategy.
Which RL algorithms are supported?
PPO (default), A2C, DQN, QRDQN, TRPO, and MaskablePPO are curated in the UI, plus an advanced model-type option for any algorithm the engine image supports.
Can I test a trained model without risking funds?
Yes. Point the model at a dry-run bot first — it trades simulated orders against live prices, so you can watch real behaviour before going live.
What plans include FreqAI cloud?
FreqAI-native reinforcement learning is available on Pro and Enterprise.
Related capabilities.
Ship your first live bot this afternoon.
Connect an exchange, build a strategy in the visual builder, backtest it on real data, and deploy. Start a 7-day Pro trial — no credit card required.
No credit card required · Cancel any time