Pricing
Pay AWS for GPU hours.Stop paying per token.
Compute runs in your own AWS account at your own rates. SynqTune does not charge a platform fee today, and you will be told before that changes.
Self-serve
Available now$0platform fee
+ AWS compute in your account, from $0.526/hour on-demand for a T4 GPU
- Pipeline builder with templates for classification, extraction, tagging, redaction and summarization
- Data preparation steps, LoRA fine-tuning and evaluation on held-out data
- Comparison report against the LLM you use today, with a ready-to-replace verdict
- Model versions, adapter and GGUF downloads, chat with your model on your own instance
- Everything runs in your AWS account; data and weights stay in your S3
Enterprise
Custom
For teams moving production traffic off an LLM API
- Help preparing data and choosing the first task to move off an LLM API
- Answers for your security review: architecture, IAM role and data flow
- Terms that match your procurement process
Calculator
What would replacing your LLM API save?
An API charges per token, so cost grows with every request. A GPU charges per hour, so at steady volume the cost per request keeps falling. Enter your traffic to see where you land.
Your current LLM usage
Example values. Replace them with your numbers and your provider's current prices.
Per month
- Current API spend
- $576.00
- GPU serving (24 h/day, on-demand)
- $724.32
- Tokens no longer sent to an API
- 1,260M
Difference
−$148.32
At this volume a dedicated GPU costs more than the API. Fewer serving hours, a smaller GPU or more traffic changes that.
GPU prices are AWS on-demand list prices in us-east-1. Spot capacity, savings plans and your own discounts lower them.
FAQ
Questions about cost
What do I actually pay for?
AWS compute and storage, billed by AWS to your own account at your rates: GPU hours while a model trains or serves, CPU hours for data preparation, and S3 storage. Spot capacity is used when available. If you turn on LLM grading or the comparison with your current LLM, that provider bills your API key for those requests.
Does SynqTune add a markup to compute?
No. Instances run in your account and appear on your AWS bill like any other workload, so your credits, savings plans and discounts apply.
Is a dedicated GPU always cheaper than an API?
No. A GPU costs the same per hour whether it serves one request or a million, so it wins at steady, high volume and loses at low volume. The calculator above shows where your break-even is. For low traffic, serve the GGUF file on hardware you already have, or run the GPU only during working hours.
How do I know the small model is good enough?
Every training pipeline can compare the fine-tuned model with the LLM you use today on the same held-out examples, scored for your task, and ends with a verdict: ready to replace, close, or not ready. You decide to switch only after seeing that report.
What does a training run cost?
The GPU hours it uses. The builder shows the instance price per hour and per day and a rough estimate for the run before you launch, and instances terminate when the run ends.