InferGrade

Recommended for you

Choose the quantized setup

Recent runs

Tracked execution

Secondary tools

Build, export, and learn

Top contributors

Community evidence should stay cumulative, exportable, and easy to inspect.

Recommendations

Which setup should I run?

Decision input

Describe the deployment shape and let InferGrade rank the current evidence.

Use case Deployment profile Execution preference Hardware class Trust preference Max VRAM (GB) Target TTFT (ms) Max benchmark cost (USD) Family filter Quant filter Capability state Table sort Plot x-axis

Decision table

Use the table first, then open the rows and families that need a closer read.

Leading candidates

Shortlist cards stay secondary to the table and plot.

Browse raw evidence

Keep one foot in the broader catalog while the decision slice stays active.

Best under VRAM

Practical local choices for the current envelope.

Pareto frontier

Trade-offs worth opening directly in compare.

Historical Results

Recent benchmark evidence

Model	Backend	Use Case	TTFT	Tok/s	Hardware	Capability	Verification

Compare

Choose between families, variants, and quants

Preset views

Start from a useful model-choice stance, then refine the exact variants or drop to raw runs.

Raw result drill-down

Result Detail

One run, fully explained

Family Explorer

Branches, quants, and nearby slices

Build

Create a short decision-suite run

Start with a local-friendly decision suite. Expand only when you need deeper reference evidence.

Benchmark identity

Name the run and the decision it should support.

Name Description Created by Tags Model Backend

Capability suites

Choose the questions this run should answer.

Benchmark groups

Broaden or narrow the evidence behind those suites.

Individual checks

These are the exact checks that will run.

Resolve from Hugging Face

Paste a repo id, model URL, or `hf://` artifact.

Hugging Face reference Artifact choice

InferGrade fills the runtime and ontology defaults.

Advanced overrides

Artifact and runtime

Only adjust these if you need an override.

Artifact URI Artifact filename Artifact SHA256 Backend image Artifact cache dir Hourly rate (USD) Notes

Ontology hints

Most users should keep the inferred values.

Family name Checkpoint name Training stage Quantization family Quantization scheme Weight bits

Generated Config

Stored server-side so local and cloud execution stay consistent.

No run config generated yet.

Execute this run

Start a tracked local or cloud run directly from the Hub.

Run locally

Pair a machine once, keep it listening, and queue tracked runs.

No local run has been created for this config yet.

Start a runner

InferGrade will highlight the one next action that matters for this config.

Advanced local commands

Start listener

Use this when no paired runner is listening.

Run immediately

Run this once instead of waiting for the listener to claim it.

Run in cloud

Create a cloud run and estimate cost for the selected provider.

Verification notes

Use a real run so deployment telemetry is measured.
Keep the artifact pinned so the bundle can be verified.
Let the run finish and upload automatically so trust labels can be applied.

Manual fallback commands

Preflight only

Use this if you want to inspect environment readiness yourself before starting a run.

Execute only

Run the config directly without the tracked local run-job wrapper.

Upload only

Publish a bundle manually if you ran outside the tracked run-job flow.

Run Status

Active and recent runs

Recent runs

Pick a run to inspect its current stage and progress.

Live timeline

Use the timeline to understand what changed and why.

Stored Run Configs

Reusable benchmark programs

My Runs

InferGrade Hub

What should happen next

Readiness and setup

Pair, start, run

Choose the quantized setup

Tracked execution

Build, export, and learn

Top contributors

Which setup should I run?

Decision input

Decision table

Leading candidates

Browse raw evidence

Best under VRAM

Pareto frontier

Recent benchmark evidence

Choose between families, variants, and quants

Preset views

One run, fully explained

Branches, quants, and nearby slices

Create a short decision-suite run

Generated Config

Execute this run

Run locally

Start a runner

Start listener

Run immediately

Run in cloud

Verification notes

Preflight only

Execute only

Upload only

Active and recent runs

Recent runs

Live timeline

Reusable benchmark programs

Contributor activity