Lanthrop
Qwen3.8-27B (4-bit NVFP4) on one RTX 5090 in California, USA. Zero data retention: no prompt or output logs, KV cache in memory only.
Lanthrop sells inference on Architect Inference. A buyer names one of its models as lanthrop/<model>, or names the model and pays the lowest offer that meets the buyer's rules.
Facts
- Handle
- lanthrop
- Website
- https://lanthrop.ai/
- Privacy policy
- https://lanthrop.ai/privacy/#api
- On the exchange since
- 2026-11-20
- Models
- 1
- Offers
- 1
- Regions
- us
- Precisions (declared)
- nvfp4
- Lowest price on
- 0 of 1 model, for the agent mix
Reliability, last 3 days
- Uptime
- 95.8% in 4 h
- Hours with an answer
- 100.0%
- Median response time
- 2.1s
Checked by the exchange
0 of 1 offer is reID verified.
The exchange verifies each offer with reID. It also benchmarks the providers.
Data handling
- Zero retention, declared by the provider
- No training, declared by the provider
Models
- Qwen 3.8 27B: us, declared precision nvfp4, $0.015 input and $0.2 output per million tokens, $0.004 cache read per million tokens, 65.5K tokens of context