Our take
Parasail is a mid-size host for downloadable models with a zero-retention data policy and measured throughput on every tracked offer. It sits between budget floors and premium tiers, with 37 open-weight models available.
Use this when zero data retention is required — it is explicitly offered. Pick it for throughput-sensitive workloads where 72 TPS on Llama 3.2 3B Instruct or 44 TPS on DeepSeek V4 Flash meets latency needs. Choose it for exploratory access to 37 open-weight models without premium pricing. Skip it if you need verified SOC 2, a confirmed EU-region endpoint, or documented headquarters location for your compliance review.
- Explicit zero-retention data policy, paired with a commitment not to train on prompts.
- Throughput transparency on every sampled offer — 12 of 12 include measured throughput, ranging from 16 to 72 TPS.
- Competitive input pricing on several models, including Mistral Nemo and Llama 3.2 3B Instruct.
- Compliance and geographic documentation is thin: SOC 2, EU endpoint and headquarters country are all unverified in our data.
- Output prices can spike relative to input on smaller models — Llama 3.2 3B Instruct output is 6.6× its input, and Gemma 4 26B A4B output is 3.1× its input.
- Some throughput figures lag for mid-size models: Qwen3.5 9B at 16 TPS, Gemma 4 26B A4B at 27 TPS and Gemma 4 31B at 28 TPS, versus 72 TPS on Llama 3.2 3B Instruct.
Privacy & data handling
- ✓
- Yes
- ✗
- No
- No answer on record
The colour says whether the answer favours you, not whether it is a yes — not training on your prompts earns a green cross, no zero-retention option earns a red one.
Prompt retention: none.