Pricing
Transparent, usage-based pricing designed to scale with your needs.
Pricing Model
Datablocks pricing consists of three components: training costs (one-time), datablock loading, and token usage.
| Component | Qwen Price | Llama Price | Notes |
|---|---|---|---|
| Training | $3.00 / 100K tokens | $2.00 / 100K tokens | One-time cost per datablock |
| Datablock Load (first) | $0.001 / load | $0.001 / load | Initial cache load |
| Datablock Load (cached) | $0.0001 / load | $0.0001 / load | Subsequent loads |
| Input Tokens | $0.60 / 1M | $0.40 / 1M | User prompts only |
| Output Tokens | $1.80 / 1M | $1.20 / 1M | Generated responses |
Cost Savings Example
Compare traditional LLM costs vs. datablocks for 1,000 queries on a 100K token document:
Traditional Approach (Qwen)
Per Query:
- 100K context tokens × $0.60/1M = $0.060
- 100 query tokens × $0.60/1M = $0.00006
- 200 output tokens × $1.80/1M = $0.00036
Total for 1,000 queries: $60.40
With Datablocks (Qwen)
One-Time Training:
- Training: $3.00 (one-time)
Per Query:
- Datablock load: ~$0.0001 (cached)
- 100 query tokens × $0.60/1M = $0.00006
- 200 output tokens × $1.80/1M = $0.00036
Total for 1,000 queries: $3.56
💰 Save $56.84 (94% cost reduction)
Break-Even Analysis
With a $3 training cost and $0.060 savings per query, you break even after just 50 queries. Every query after that represents pure cost savings.
Plans
Free Trial
$0
- ✓1M free tokens
- ✓Train up to 2 datablocks
- ✓Full API access
- ✓Community support
Pricing FAQ
How is training charged?
Training is a one-time cost based on the size of your document. Once trained, you can use the datablock unlimited times without retraining.
What happens if my datablock gets evicted from cache?
You'll pay the first-load price ($0.001) to reload it. Frequently used datablocks stay cached longer.
Can I delete datablocks I no longer need?
Yes, you can delete datablocks anytime through the dashboard or API. Deleted datablocks won't incur any ongoing costs.