VIDRAFT · Unified Inference Engine
VKAE accelerates. VKUE saves. VKIE maximizes serving. One 34.7B model, from a datacenter GPU down to a free CPU — every number measured, every demo live.
Speed, cost, and scale pull against each other — win one, usually lose two.
Slow inference is unusable — but going fast normally means an expensive datacenter GPU.
Serving a large model means a multi-GPU cluster. With GPU shortages, you often can't even buy them.
Model and hardware get locked in — no single engine spans datacenter down to edge / on-prem.
GPU acceleration · the sports car
GPU savings · the compact car
accel + savings = max serving · the train
Each axis, on the same hardware, same 34.7B model — what our engine changes.
Same 34.7B model, same principles — three optimization targets.
| 🏎️ VKAE | 🚗 VKUE | 🚄 VKIE | |
|---|---|---|---|
| Focus | Speed | Savings | Unified · throughput |
| Best hardware | Datacenter GPU | CPU ~ small GPU | Full range |
| Strength | Top single-stream speed | Lowest cost · accessibility | Max tok/s · cost-efficiency |
| Measured before→after | 24→220 (9×) | 5.4→20 (3.7×) | 24→18,057 (750×) |
| In a phrase | fastest | cheapest | most |
Ourbox-35B-JGOS — 34.7B total / ~3B active MoE. Same weights, only the hardware changes.
| Hardware | Measured tok/s | Axis |
|---|---|---|
| 1× B200 (datacenter) | 18,057 | VKIE ceiling · aggregate (256 concurrent) |
| 1× A10G (cloud GPU) | 126 | VKUE · single-stream |
| 8 GB gaming laptop (RTX 5060) | 20.0 | VKUE · 3.7× a dense 32B |
| CPU-Upgrade (8 vCPU, no GPU) | ~17 | VKUE |
| FREE CPU space (2 vCPU, no GPU) | ~6-7 | VKUE floor · zero cost |
Quality holds across every tier — GPQA Diamond 86.4% (Ourbox-35B, maj@8) up to 90.9% (Darwin-398B). Multimodal: Janus-Pro-1B image generation on a low-cost T4 in ~28 s/image (fp16, 2.1× our first cut).
Breaking the rule that "big AI needs big money."
Public sector, defense, healthcare, finance — data that can't touch the cloud. Frontier reasoning on an air-gapped on-prem CPU.
From a multi-hundred-thousand-dollar GPU cluster to a ~$1,600 card — or a free CPU. Entry cost drops by orders of magnitude.
Individuals, startups, SMBs, public bodies — anyone. Ready for the surge in on-device and edge demand.
These run on their real hardware (not this page). Click a card to open, or load one on screen below.