Our take
Wafer is a host for downloadable models whose cheapest rows are its flash-tier DeepSeek offers, and it states plainly that it does not train on prompts. Its compliance posture and location are unverified in our data, so regulated work needs confirming elsewhere.
Use it for cost-sensitive work on flash-tier models, where its DeepSeek rows are the cheapest it lists, or where a zero-retention option matters and prompts must not be used for training. Skip it if you need a verified security attestation or a guaranteed European endpoint, or if you need measured quality or speed before committing.
- Its flash-tier DeepSeek rows are the cheapest it lists, well under its own GLM 5.3 rate, so that tier is where the low rates sit.
- It does not train on prompts and offers a zero-retention option, which gives sensitive work a route that does not feed the model.
- GLM 5.3 appears twice at different input rates, so check both rows before assuming one price for that model.
- An independent security attestation and a European endpoint are both unverified in our data, so a compliance or data-residency requirement cannot be confirmed from what we list.
- No benchmark score and no throughput figure for any of its offers, so quality and speed cannot be compared on evidence here.
Point your tools here
We have not recorded what a router needs for Wafer yet — no base URL and no docs link on file. That is our gap, not a sign Wafer has no API; its own documentation is the place to look until we close it.
Privacy & data handling
These answers cover requests sent to Wafer through OpenRouter, as OpenRouter records them, last read 4 hours ago. For Wafer's own API we hold no answer: check its terms. SOC 2 and the data processing agreement below are hand-checked, not part of that sync.
Prompt retention: none.