Doppler

Text Generation

Output

Idle

Relative perplexity low med high

Proof Console

Side-by-side translation compare

Hold one source sentence constant, then compare engine, model, size, and latency receipts without leaving the page.

Evidence bundle Awaiting frozen scoreboard Gamma artifacts and eval rows attach here.

Teacher vs student Delta pending BLEU / chrF / size deltas render once evidence arrives.

Public claim Practical EN/ES parity test Narrow proof scope, explicit caveats, no hidden boosts.

Show benchmark receipts

Artifact id Pending

Browser / device Awaiting run

Saved receipts 0

No benchmark receipts linked yet.

Evidence snapshot No receipt loaded

Teacher BLEU --

Student BLEU --

Teacher chrF --

Student chrF --

Size delta --

Artifact link Pending

Sample panel

Deterministic smoke panel

3 easy · 3 nuanced · 2 domain · 2 edge cases

Left

Reference lane

Pin the baseline you want the audience to trust first.

baseline

Engine

Model

Status Idle

Model size--

Load--

First token--

Decode--

Total--

DeviceWebGPU

Output No run yet

Awaiting compare run.

Right

Challenger lane

Swap student, engine, or both without changing the source request.

student

Engine

Model

Status Idle

Model size--

Load--

First token--

Decode--

Total--

DeviceWebGPU

Output No run yet

Awaiting compare run.

Stacked history

Live proof log

Pending sample

Compare receipts will stack here with engine/model labels, timing, and expandable outputs.

en -> es No run captured

Awaiting artifact

Share links, receipt ids, and lane winners can render as badges instead of replacing raw numbers.

TTFT / decode / total Transparent first

Image Generation

Output

Idle

Energy Optimization

Idle

Board (thresholded)

Vector (thresholded)

Board (intensity)

Energy over time

Stats

Steps --

Energy --

Symmetry --

Count --

Binarize --

Runtime --

Diagnostics

Run Profile

Idle

No output yet.

Raw JSON

No JSON yet.

Advanced

Runtime preset (advanced override)

Runtime override (JSON file)

Preset: modes/debug

Runtime override preset

Tooling intent

Model Intake

Ready

Import from URL(s) (GGUF/safetensors, one per line)

Import pre-converted RDRR from base URL

RDRR base URL

Advanced

Model ID override (optional)

Model preset override (optional)

Weight dtype override (optional)

Energy Controls

Energy demo

Advanced

Quintel

Size

Display threshold

Count target

Rules

Mirror X Mirror Y Diagonal Count

Weights

Symmetry

Count

Binarize

Init

Mode

Seed

Scale

Loop

Steps

Step size

Gradient scale

Convergence

Models & Storage

Storage

OPFS usage -- / --

GPU Info

Device --

Buffer Limit --

Features --

Diagnostics Stats

GPU Memory (tracked)

Tracked GPU --

Peak GPU --

Buffers --

Requested --

Pool hit rate --

Max buffer --

Top buffers (tracked) [?]

No tracked buffers yet.

JS / CPU

JS Heap --

System RAM (est) --

Swap Not available

Storage (OPFS)

OPFS usage --

Active model --

Memory Control

Run Controls

Prompt

Generation settings

Temperature

auto: --

Top P

auto: --

Top K

auto: --

Max tokens

auto: --

Reset context each run

Token Press

Diffusion Controls

Prompt

0/2048

Advanced

Negative prompt

0/1024

Steps

Guidance

Seed

Width

Height

Inference Pulse

Performance

Tokens/sec --

TTFT --

Prefill --

End-to-end --

Decode --

Prompt / Gen --

Recent Runs

Run TTFT Prefill Decode E2E

KV Cache

Allocated --

Used --

Efficiency --

Seq / Max --

Layout --

Text Generation

Side-by-side translation compare

Deterministic smoke panel

Reference lane

Challenger lane

Image Generation

Energy Optimization

Stats

Diagnostics

Model Intake

Energy Controls

Models & Storage

Storage

GPU Info

Diagnostics Stats

GPU Memory (tracked)

JS / CPU

Storage (OPFS)

Memory Control

Run Controls

Diffusion Controls

Inference Pulse

Performance

Recent Runs

KV Cache

Error