Free beta // until 2026-09-16
Unfiltered
intelligence.
A private-by-default inference layer for builders. Qwen3.8 27B OBLITERATED, IQ4, DFlash2 speculative decoding, and vision — served by Venym Labs hardware.
Model manifest
0 nodes seenQwen3.8 27B OBLITERATED
TARGET
IQ4_XS
DRAFT
DFlash2
VISION
READY
No training. No prompt or response retention. Usage metadata only for service operation and billing.
01 // Playground
Talk to the model.
sign in to run
QWEN3.8 27B · OBLITERATED · IQ4 · DFLASH2
EPHEMERAL SESSIONStart with a sharp question.
This thread lives in your browser only. Refreshing clears it. Nothing is written to the inference service.
MULTI-NODE ROUTING
Every request can move across healthy Vast nodes.
NO DATA PRODUCT
No training corpus. No conversation archive.
OPENAI-COMPATIBLE
One endpoint for apps, agents, and local tooling.