❓Fast Worker FAQ

Sign in at dashboard.sogni.ai/fast-workers. The Workers page shows every Fast and Relaxed worker on the account, downtime alerts, live status, jobs, health, settings supported by the worker, and earnings analytics.

See Using the Workers Dashboard for a guided tour.

An Online indicator confirms the worker is connected. Select the worker to inspect its current model, active job, last job, loaded models, recent failures, health telemetry, and job history.

An online worker is not guaranteed to have an active job. Demand, model eligibility, cached workflows, speed, and other available workers all affect assignment. If the worker is online but idle, first check model-download progress and recent errors rather than repeatedly restarting it.

The dashboard shows job history, payment types, estimated values, and recent transactions. Regular Spark- and SOGNI-paid jobs credit the worker under the normal paid-job path.

Unlimited subscription-covered jobs are optional and use the monthly Worker Subscription Revenue Share Pool. Review participation and pool accounting at dashboard.sogni.ai/subscription-earnings. For finalized earnings, follow Receive subscription earnings in USDC to set up your payout wallet on Base.

Demand and earnings vary. Dashboard examples and historical leaderboards are not guarantees of future workload, revenue, profit, or operating-cost recovery.

Read What an Nvidia GPU earns on Sogni in Sept 2026 for GPU-by-GPU September scenarios grounded in production worker activity. The article covers pay-as-you-go earnings and the Unlimited subscription pool, with explicit assumptions about continuous availability, workload share and further pool growth.

These are dated forecasts before electricity, hosting and hardware costs. Use your own dashboard results to assess your setup; the article is not a guaranteed rate or an estimate for Mac or LLM workers.

Yes, when an online Sogni Comfy Worker supports remote configuration. The Settings tab first appeared with Comfy Worker 1.0.189; use the latest supported release for the current settings catalog and live-apply behavior. Open the worker's Settings tab to configure every operator-tunable environment setting the worker publishes, see each value's effective source, and check whether a change is live or requires a restart. Credentials, worker identity, and provider secrets remain deployment-managed.

Stable Diffusion Worker and LLM Worker settings remain deployment-managed. Compatible workers can also receive accepted dashboard overrides by Fast Worker NFT token ID when they reconnect without local overrides. See Advanced Configuration for precedence, recovery, and persistence.

No. Current Comfy Workers provide best-effort render-time rests, GPU telemetry, and a temperature-based guard. By default, the guard stops admitting new work at 188.6°F (87°C) and can resume at 174.2°F (79°C), after at least a 60-second hold. It does not cancel an active render, and it cannot help when sensor data is unavailable.

These software features are not a safety certification or a substitute for correct physical cooling, airflow, power delivery, ambient conditions, maintenance, vendor limits, and independent monitoring. The owner/operator is responsible for deciding whether the system and settings are safe. Read Protecting your hardware.

Fast Worker requires a supported NVIDIA GPU. Some current workflows begin at 16 GB of VRAM, while others require 20 GB, 24 GB, 32 GB, or more. GPU architecture, driver/runtime compatibility, system memory, and configured model policy also affect eligibility.

Because the catalog changes, do not use an old static GPU/model list for a hardware purchase. Check the live Comfy workflow catalog, the Stable Diffusion model catalog, and the current installer. A supported GPU is eligible only for the workloads whose current requirements it meets.

Create or sign in to your Sogni account, then run the current worker installer. Its authentication flow guides you through the account API key and worker NFT setup.

Treat the API key as a secret. Each concurrently running GPU worker needs its own NFT token ID. Workers owned by the same account can use the same account API key, but must not run with the same NFT token ID.

Yes. Use one primary Sogni account, one API key for that account, and a unique Fast Worker NFT token ID for each concurrently running GPU worker.

On a hosted platform, keep each deployment at one replica unless every replica receives a different NFT token ID. On a local multi-GPU host, use the installer's multi-GPU generator. Size host RAM, storage bandwidth, Internet, power, airflow, and cooling for the combined load—not for one GPU in isolation.

Yes. The Workers dashboard groups both networks under the account and applies account-wide downtime-alert settings to both. Each worker still needs the identity and setup required by its own worker type.

Yes. Run the current installer again and choose the new worker type. Preserve the worker data and model directories unless you intentionally want to discard them, and verify the new type and version in the dashboard afterward.

Flux Worker is retired and must not be selected or reactivated. Use Comfy Worker for current Flux-family workflows.

Update available means a newer supported worker image is ready. A compatible online worker may offer Upgrade worker; after confirmation, it finishes its current job, pulls the new image, recreates the worker, and reconnects. If the banner offers Update instructions, use the manual installer/update process from the worker's installation directory.

Review the Fast Worker Release Notes to see what changed, whether the update is required or recommended, and the verified image digests.

Upgrade required means the running version is too old for a required network or dashboard capability. Update it before troubleshooting a missing Settings tab or new workload. After updating, confirm the reported version and a successful reconnect at dashboard.sogni.ai/fast-workers.

Custom hosted deployments pinned to an immutable image do not automatically follow a latest tag. Query the current official image endpoints and update the provider deployment deliberately.

At startup, Comfy Worker reads the production workflow catalog, filters it for the detected GPU and operator policy, verifies the persistent cache, and queues missing dependencies for eligible workflows. Shared model components are downloaded once even when several workflows use them. The worker can connect after usable work is ready and continue preparing other eligible files in the background.

A normal restart scans the cache again and skips verified files. A download can remain incomplete because of a storage cap, insufficient free space, repeated network/integrity failures, or an interrupted/ephemeral deployment.

The full eligible cache can consume several hundred gigabytes and changes with the catalog. Use PREFERRED_WORKFLOW_IDS, download controls, and storage limits in Advanced Configuration instead of relying on a dated fixed total.

Light use is possible, but games, local AI tools, video work, sleep/hibernation, or other GPU-heavy applications can slow jobs, consume VRAM, cause failures, or reduce assignment priority. Stop the worker before intensive GPU use and start it again afterward.

Remember that the worker's temperature guard does not manage heat or power created by unrelated applications.

Use the NFT management interface at nft.sogni.ai if the token is eligible for transfer. Stop the worker before transferring it. The new owner must configure the worker with that account's API key before starting it again.

Dashboard settings recovery is keyed to the NFT token ID. Review and clear dashboard overrides before transfer if machine-specific thermal, workload, or download values should not follow the NFT to a compatible replacement worker.

NFT-level history and program treatment can be subject to current product rules and terms. Review the destination account and displayed transfer details before confirming.

Yes. Eligibility and blocked locations can change with sanctions, export controls, provider policy, and other legal requirements. The Sogni Terms of Service and the platform's current enforcement are controlling; a static country list in documentation can become inaccurate.

Do not use a cloud provider, proxy, routing choice, or false account information to evade a restriction. The operator is responsible for local law, taxes, electricity and hosting arrangements, and provider terms.

Start with the worker's Failed Jobs & Errors, Worker Health, and Worker Care panels and the local worker-debug script. Do not publish an API key or copied runtime logs before reviewing them.

For support, email [email protected] or join the Sogni Discord.