Shamal AI
Drone Intelligence Dashboard
CHECKING…
LIVE
◐
Detections
—
Target class
—
Latency
—
Tokens · in / out
—
Est. cost
—
Parameters
CONFIG
Models
VLM · reasoning
Qwen3-VL-30B-A3B-Instruct · Novita
Qwen2.5-VL-72B-Instruct
Qwen2.5-VL-7B-Instruct
Detector
Falcon Perception 0.6B
Detection task
Segment
Detect
Inference
VLM temperature
0.10
VLM max tokens
512
Falcon max tokens
100
Falcon temperature
0.00
Max objects
8
Confidence ≥
0.35
Compute dtype
bfloat16
float32
Seed
Deployment
VLM provider
Novita / HF
Detector
ZeroGPU · HF
Data residency
In-Kingdom
Languages
EN · AR
Price $/1M tok · in / out
Price is an editable estimate for the cost readout — confirm Novita / HF rates before quoting.
Imagery
CONSTRUCTION
·
—
Drop an aerial / drone frame, click to upload,
or load a demo frame.
JPG · PNG · top-down imagery works best
Run
Demo frame
Upload
Sample frames · click to load
Detection · Falcon
PERCEPTION 0.6B
Run a query to detect objects on the frame.
Reasoning · Qwen
QWEN3-VL
Grounded natural-language answer appears here.
Run output
JSON
{ }