Comfy Worker Release Notes

The public history of Sogni Comfy Worker releases. Each entry says whether to update, what changed for your worker, and whether it needs disk space, a setting, or a driver.

#Updating

• Installer-managed workers follow the latest image. Use Upgrade worker in the Dashboard, or run worker-restart.bat on Windows or ./worker-restart.sh on Linux. The background updater waits until the worker is idle.
• A deployment pinned to a numbered tag (for example sogni/comfy-worker-cu13:1.0.206) must move the tag by hand.
• Do not interrupt active work. Prefer the managed upgrade or wait for an idle moment.
• Use the CUDA 13 image. It needs an NVIDIA RTX 30-series or newer GPU with driver R580 or newer, and installer 1.0.181 and newer run only this image. The CUDA 12 image is legacy: it is still published for existing deployments until they move to CUDA 13, and it will be retired. Do not use it for a new worker.

dashboard.sogni.ai/fast-workers shows the version each worker runs and whether an update is available. The Windows and Linux scripts have their own installer release notes.

#Version 1.0.248

Released: October 28, 2026 PT · Update: Recommended for CUDA 13 workers.

• Image previews consistently show the current render.
• Delayed preview updates stay with the generation that produced them.

Disk and settings: Small automatic download; no additional model weights or worker settings.

#Version 1.0.247

Released: October 27, 2026 PT · Update: Recommended for CUDA 13 workers.

• Wan Animate jobs release finished frames, masks and video data as soon as the job no longer needs them, so long Animate videos use much less system memory.
• Wan 2.2 jobs use less system memory and return it after the model unloads, so Wan video now runs on workers with 32 GB of RAM.

Disk and settings: Small automatic download; no additional model weights or worker settings.

#Version 1.0.246

Released: October 27, 2026 PT · Update: Recommended for CUDA 13 workers.

• Downloads large model files faster on fast connections by fetching several parts of each file at once. Slow connections download at the same speed as before.
• An interrupted download still resumes where it stopped, and every file is still checked before the worker uses it.

Disk and settings: Small automatic download; no additional model weights or worker settings.

#Version 1.0.245

Released: October 26, 2026 PT · Update: Recommended for CUDA 13 workers.

• Supports optional reference-video compression for one-stage MiniMax H3 Standard, Balanced, LightX2V Turbo and FastH3 jobs.
• Allows tested Balanced reference-video outputs up to eight seconds and Standard outputs up to 5.17 seconds at 1344×768 on a 32 GB RTX 5090. Existing jobs keep their normal behavior.

Disk and settings: Small automatic download; no additional model weights or worker settings.

#Version 1.0.244

Released: October 24, 2026 PT · Update: Recommended for CUDA 13 workers.

• Improves image checks and checks finished videos when the Sensitive Content Filter is on.
• Fixes color shifts in generated videos and MiniMax H3 extensions.

Disk and settings: Small automatic download; no configuration changes.

#Version 1.0.243

Released: October 23, 2026 PT · Update: Recommended for CUDA 13 workers.

• Keeps taking jobs when Docker Desktop on Windows (WSL2) cannot lock more memory for fast GPU transfers. Earlier workers restarted instead, and some hosts then downloaded their models again.
• Locks at most 45% of system memory for fast GPU transfers under WSL2, the same limit as on Windows. Earlier workers could try to lock 95% of it.

Disk and settings: No new downloads or settings.

#Version 1.0.242

Released: October 23, 2026 PT · Update: Recommended for CUDA 13 workers.

• Keeps taking jobs when an update has to wait its turn. Earlier workers could stop taking jobs for up to an hour after their update was postponed.
• Keeps the sound continuous where a MiniMax H3 extension joins the original video. Earlier extensions could fall silent for about 40 ms at the join.
• Fits MiniMax H3 reference audio shorter than 2 seconds or longer than 15 seconds instead of failing the render.
• Uses LoRAs a worker already has when its models folder cannot be written. Earlier workers failed those jobs.

Disk and settings: No new downloads or settings.

#Version 1.0.241

Released: October 22, 2026 PT · Update: Recommended for CUDA 13 workers.

• Keeps every reference photo whole in Qwen Image Edit 2511 and its Lightning version, which cropped references to the output's shape and could cut off a face.
• Sizes image edit references to the output instead of cropping them, at most 2 MiB each.
• Updates a worker as soon as its current job finishes. Earlier workers also waited for background model downloads, which could keep a worker out of work for up to an hour per update.

Disk and settings: No new downloads or settings.

#Version 1.0.240

Released: October 21, 2026 PT · Update: Recommended for CUDA 13 workers.

• Extends MiniMax H3 videos with two-stage FastH3 and three or more LoRAs, which earlier workers could not build.
• Fixes MiniMax H3 Reference to Video failing when a reference video with an odd width or height had to be retimed.
• Keeps MiniMax H3 Reference to Video at 1080p and 2K working with LoRAs on every tier, including FastH3 Reference to Video.
• Applies MiniMax H3 LoRAs that train the output layer to FastH3 Reference to Video, which refused or partly skipped them before.

Disk and settings: No new downloads or settings.

#Version 1.0.239

Released: October 21, 2026 PT · Update: Recommended for CUDA 13 workers.

• Adds FastH3 Reference to Video on CUDA 13 workers: a fast MiniMax H3 model that makes video from reference pictures, clips and sound.
• Gives long video renders enough time for their full size and length, so slow GPUs no longer stop long Wan 2.2 and MiniMax H3 clips as timed out while they are still rendering, and still stops a Wan 2.2 render that has truly stalled.
• Recovers from a stuck render, or from a cancelled job that leaves the GPU busy, by restarting only the render engine, so the worker keeps its downloaded models and is back in seconds.
• Applies Krea 2 LoRAs exported by fal's trainer, which the worker previously could not use.
• Releases pinned system memory correctly after memory pinning switches itself off, so that memory is no longer held for the rest of the run.

Disk and settings: FastH3 Reference to Video downloads about 22 GB on CUDA 13 workers with at least 32 GB VRAM. No new settings.

#Version 1.0.238

Released: October 20, 2026 PT · Update: Recommended for CUDA 13 workers.

• Supports MiniMax H3 video extensions: continues a video from where it ends, carrying its last second of motion and sound into the new footage.
• Scales very large Reference to Video pictures down before loading them, so oversized photos no longer cost extra memory or time.

Disk and settings: No new downloads or settings.

#Version 1.0.237

Released: October 19, 2026 PT · Update: Recommended for CUDA 13 workers.

• Supports Mystic X v5 and the Action Speed Slider for MiniMax H3, including negative and higher LoRA strengths.
• Preserves negative strengths for MOSS Image LoRAs instead of turning their effect off.
• Adds support for optional MOSS composition controls.
• Avoids false job timeouts while large models load or unload.

Disk and settings: No new base-model downloads or settings; selected LoRAs download when requested.

#Version 1.0.236

Released: October 18, 2026 PT · Update: Recommended for CUDA 13 workers. Replaces the withdrawn 1.0.235.

• Adds MOSS-Image SFT, an uncensored text-to-image model, with a standard preset, an eight-step preset and Personal LoRAs on compatible workers.
• Fixes startup when a selected model uses an adapter stored in a subfolder.
• LTX video edits accept silent source videos and retain audio from sound-bearing inputs.

Disk and settings: MOSS downloads about 17 GB on CUDA 13 workers with at least 24 GB VRAM. No new settings.

Application update 1.0.236-runtime.1 — withdrawn October 18, 2026: The rollout was stopped after a canary stopped responding while switching models. Managed workers remain on 1.0.236.

#Version 1.0.235

Released: October 18, 2026 PT · Update: Withdrawn. Use 1.0.236 or the current recommended release.

• Withdrawn during its first test rollout because the worker could exit during startup. No fleet rollout followed.

Disk and settings: MOSS downloads about 17 GB on CUDA 13 workers with at least 24 GB VRAM. No new settings.

#Version 1.0.234

Released: October 16, 2026 PT · Update: Rolling out gradually; no action needed. Replaces 1.0.232.

• Adds FLUX 3 Image Klein, Black Forest Labs' fast open image model: pictures up to 4K, edits with up to ten reference photos, and artists' own LoRAs.
• Image jobs that start from an existing image now run only the steps their strength calls for, so light touch-ups finish several times faster.
• A repeat of a job the worker already ran now renders again instead of returning the earlier, unpaid result.
• MiniMax video, Qwen Image and music jobs use less temporary memory.
• Mask jobs on sideways phone photos no longer fail with a size mismatch.
• A worker stuck restarting after a crash now recovers within about 10 seconds.
• A worker over its model-folder size limit still checks the models it already has.

Disk and settings: a new base image: one image download about the size of the current one, then later updates are small again. CUDA 13 workers with 16 GB or more also download FLUX 3 Image Klein, about 11 GB; its 4K pictures go only to cards with 30 GB or more. No new settings.

#Version 1.0.233

Released: October 15, 2026 PT · Update: Superseded. Update to the current release.

• Withdrawn during its first test rollout: some image jobs that start from an existing image failed. Its changes, with that fix, ship in 1.0.234.

#Version 1.0.232

Released: October 5, 2026 PT · Update: Superseded by 1.0.233, which carries all of its changes; it reached only a few workers. No action needed.

• Short speech results, such as a one-word line, no longer fail as a corrupt file.
• LTX 2.3 videos at 4K, and long LTX 2.3 videos at high resolution, no longer fail before the first frame.
• Image-to-image jobs at a very high strength with many steps no longer fail.
• A worker that started offline from its built-in release now moves to the current release once it reaches Sogni again.

Disk and settings: a new base image, so a worker moves to it with one image download about the size of the current image; later updates are small again. No new settings.

#Version 1.0.231

Released: September 30, 2026 PT · Update: Recommended for all workers. Replaces 1.0.230.

• Workers that follow the latest image switch once to a self-updating image: later versions arrive as a download of about 230 MB and apply between jobs, with no new Docker image to pull.
• A slow or dropped connection during that download now resumes where it stopped instead of starting over.
• MiniMax video jobs no longer fail at the final decode on cards that keep part of the model in system memory.
• Finished images are now scored by the image check in the background; delivery never waits for it.
• Multi-GPU installs created before installer 1.0.179 need one installer update before they can switch: run ./worker-start.sh in a terminal and accept the update. Until then the worker's log says so.

Disk and settings: one image download when a worker switches, smaller than the current image; later updates download only about 230 MB. No new settings.

#Version 1.0.230

Released: September 29, 2026 PT · Update: Recommended for all workers; strongly recommended for MiniMax video workers. Replaces 1.0.229.

• MiniMax first-and-last-frame videos made with a LoRA now use the same attention setting as image-to-video, which keeps fast motion from smearing.
• The check that screens finished images is more accurate and about three times faster. It uses a new model that the worker downloads once, in the background, when it starts; renders are never held up while it downloads.
• The worker image is about 300 MB smaller.

Disk and settings: one new download of about 104 MB (the image-check model, stored under models/safety on the shared volume); no new settings.

#Version 1.0.229

Released: September 28, 2026 PT · Update: Recommended for all workers; strongly recommended for MiniMax video workers. Replaces 1.0.228.

• MiniMax image-to-video at 2K without LoRAs now uses the same attention setting as LoRA jobs, which keeps fast motion from smearing and finishes a little faster.
• A render that comes out broken (invalid values or a blank video) now fails and is retried on another worker instead of being delivered.
• A worker whose startup stalls now exits so it is restarted, instead of staying down.

Disk and settings: no new downloads and no new settings.

#Version 1.0.228

Released: September 27, 2026 PT · Update: Recommended for all workers; strongly recommended for MiniMax video workers. Replaces 1.0.227.

• MiniMax videos with LoRAs at 2K now apply the LoRAs to the first pass only, so the upscale stays sharp and finishes a little faster.
• MiniMax videos with LoRAs keep the attention setting that stops smearing in fast motion.

Disk and settings: no new downloads and no new settings.

#Version 1.0.227

Released: September 26, 2026 PT · Update: Recommended for all workers; strongly recommended for MiniMax video workers on machines with about 40 GB of system memory or less. Replaces 1.0.226.

• Starting a new MiniMax video prompt no longer gets the worker stopped for running out of system memory on machines with about 32 to 40 GB of RAM.
• When system memory is short, the worker now frees a model it is setting aside instead of copying it into memory first, so switching models no longer briefly needs room for both.
• A worker no longer holds earlier results in memory while it starts a new job.

Disk and settings: no new downloads and no new settings.

#Version 1.0.226

Released: September 26, 2026 PT · Update: Recommended for all workers; strongly recommended for machines with about 32 GB of system memory and for MiniMax video workers. Replaces 1.0.225.

• Krea 2 jobs with LoRAs no longer run up to three times slower right after another model on machines with about 32 GB of memory.
• Cancelling a job no longer restarts workers on machines with limited memory, and a cancelled MiniMax video now stops right away.
• MiniMax videos with LoRAs render about 30% faster on GPUs with 80 GB or more of memory, with the same output.
• Workers can now serve MiniMax videos that use intermediate keyframes.
• A worker whose container software was changed after installation now stops taking jobs until it is recreated from the published image.

Disk and settings: no new downloads and no new settings.

#Version 1.0.225

Released: September 25, 2026 PT · Update: Recommended for all workers; required for workers serving MiniMax video with LoRAs on GPUs with 80 GB or more. Replaces 1.0.224.

• MiniMax videos that use LoRAs are sharp again on GPUs with 80 GB or more of memory; those renders take somewhat longer than before.
• MiniMax videos no longer fail on machines where a package install inside the container replaced a required component with an older version.
• Jobs that use LoRAs no longer occasionally fail to start while another job finishes downloading the same LoRA, a regression in 1.0.224.
• Large finished videos no longer fail to upload on slower connections: the upload time limit now grows with the file size.
• Upscaled videos are saved at higher quality.

Disk and settings: no new downloads and no new settings.

#Version 1.0.224

Released: September 24, 2026 PT · Update: Required for workers serving Krea 2; recommended for all workers. Replaces 1.0.223.

• Krea 2 renders that use LoRAs no longer come back blank, a regression in 1.0.223.
• A GPU that stays too hot to take work while idle now stops the worker with a clear fault instead of leaving it online without jobs; after the restart the worker waits for the card to cool before it rejoins.
• Workers delete leftover job files, warmup renders and interrupted downloads, including at startup, so disk use no longer creeps up between restarts.
• A crash no longer writes a multi-gigabyte crash dump to the host's disk.

Disk and settings: no new downloads. New setting: Guard fault timeout (minutes), default 10, sets how long an idle GPU may stay above its resume temperature before the worker stops.

#Version 1.0.223

Released: September 23, 2026 PT · Update: Required for workers serving LTX or WAN Animate; recommended for MiniMax video workers.

• MiniMax videos finish decoding faster and use less GPU memory during decoding.
• MiniMax videos no longer show the circular rainbow-grid artifact corrected in this release.
• LTX output video decoding uses less memory, preventing the reproduced long-video memory crash.
• WAN Animate fixes a known CUDA 12 processing crash and uses less temporary memory for longer videos.
• Image jobs respect output-format and metadata settings more reliably, and cancelled jobs stop publishing late results.

Disk and settings: MiniMax workers download a 3.2 GB replacement and remove the previous 5.2 GB file, reclaiming about 2 GB; no new settings.

#Version 1.0.222

Released: September 21, 2026 PT · Update: Recommended for all workers, especially workers serving WAN Animate Move and Replace.

• WAN Animate Move and Replace now report realistic initial completion estimates, improving customer countdowns and queue wait estimates.
• Longer WAN Animate jobs retain their rendering allowance, so the corrected estimate does not cause them to be stopped too early.
• Disk and settings: nothing new for workers already on 1.0.221.

#Version 1.0.221

Released: September 21, 2026 PT · Update: Recommended for all workers. MiniMax H3 workers should update.

• MiniMax H3 FastH3 videos made from a starting image, or from a first and last image, no longer show bright lamps, bulbs and ceiling lights wobbling or undulating, at normal size and in 2K.
• Turning on the mature-theme model opt-out while the worker runs now stops those downloads straight away, instead of finishing the pack first.
• Disk and settings: nothing new for workers already on 1.0.220.

#Version 1.0.220

Released: September 18, 2026 PT · Update: Recommended for all workers. MiniMax H3 workers need it to serve Reference-to-Video in 2K (and 1080p).

• Reference-to-Video in 2K. MiniMax H3 two-stage reference-to-video now delivers 2688×1536 (and 1920×1088) from up to 9 reference images, 3 reference videos, and 3 audio clips, and every two-stage reference-to-video clip length renders on 32 GB cards: the worker renders the reference pass at the 768p canvas on the tier's own model, then hands the result to the FastH3 refinement stage for the larger sizes. Model IDs minimax-h3-ref2va-fp8_r2v_2stage and _r2v_balanced_2stage with the 1344×768 (or portrait) canvas.
• The 1080p and 2K reference-to-video classes go only to CUDA 13 workers that carry the FastH3 checkpoint and have at least 60 GB of host RAM; 768p keeps the 1.0.218 rules.
• Disk and settings: nothing new for workers already on 1.0.218.
• 1.0.219 was published with the same changes but never recommended: its first two-stage reference-to-video job failed at graph build, and 1.0.220 carries the fix. Move any 1.0.219 worker to 1.0.220.

#Version 1.0.218

Released: September 17, 2026 PT · Update: Recommended for all workers. MiniMax H3 workers need it for the new two-stage reference-to-video jobs.

• Oversized uploads no longer crash the worker: reference videos and images are shrunk to what the job needs before rendering, which ends the out-of-memory restarts that video-driven jobs (animate, video-to-video, reference-to-video) caused on many hosts.
• MiniMax H3 two-stage output no longer shows the warm glowing patch that could appear in small areas such as clothing.
• The same FastH3 job now renders identically whether it is the first job after startup or follows another, so results are reproducible from the same seed.
• Two-stage reference-to-video jobs go only to 1.0.218 or newer.
• On 24 GB and 32 GB cards, Standard and Balanced MiniMax H3 jobs now go only to hosts with at least 39 GB of RAM; hosts with less kept crashing on them.
• Disk and settings: nothing new for workers already on 1.0.217.

#Version 1.0.217

Released: September 14, 2026 PT · Update: Recommended for all workers. MiniMax H3 and Pixal3D workers need it for the new jobs below.

• MiniMax H3 workers can take video jobs driven by an uploaded audio track, from a photo, from first and last frames, or from audio alone; these go only to 1.0.217 or newer.
• Pixal3D workers can build a 3D model from up to four photos of an object; these jobs go only to 1.0.217 or newer.
• MiniMax H3 2K output no longer shows the faint specks and corner lines 1.0.216 left along tile seams.
• Disk and settings: a worker with the Pixal3D pack downloads one more 5.2 GB model on restart. No new settings.

#Version 1.0.216

Released: September 13, 2026 PT · Update: Recommended for all workers.

• MiniMax H3 2K output has far fewer streaks along tile seams.
• A Fast worker with less than 31 GB of RAM now stays disconnected and reports itself unhealthy so its host can replace it, instead of reconnecting over and over. Relaxed workers are unchanged.
• Disk and settings: nothing new for workers already on 1.0.215.

#Version 1.0.215

Released: September 12, 2026 PT · Update: Recommended for all workers. Replaces 1.0.214, which was never made current.

• Renders that stack several personal LoRAs go only to workers on this version or newer.
• Future changes to personal LoRA rules (strengths, trigger phrases, stack size) no longer need a worker update.
• When a model download would run out of disk, the worker deletes idle personal LoRAs to make room instead of failing.
• Jobs that need a specific model file go only to workers that already hold it.
• Disk and settings: a MiniMax H3 worker that has not restarted since September 12 downloads one more 659 MB file on restart. No new settings.

#Version 1.0.214

Released: September 11, 2026 PT · Update: Superseded by 1.0.215 before it became current. Update to 1.0.215 instead.

• Prepared workers for the full personal LoRA strength range. Its changes ship in 1.0.215.

#Version 1.0.213

Released: September 11, 2026 PT · Update: Recommended for all workers. Required for MiniMax H3 workers that will serve Balanced jobs with a custom video shift.

• MiniMax H3 Balanced runs one recipe in every mode and accepts a custom video shift. Jobs with a custom shift go only to 1.0.213 or newer.
• Disk and settings: frees about 2 GB by removing a superseded MiniMax H3 LoRA at startup. A few H3 workers download a 780 MB adapter they did not already hold. No new settings.

#Version 1.0.212

Released: September 11, 2026 PT · Update: Recommended for all workers. Required for personal LoRA jobs, FlashVSR jobs with non-default options, and LTX audio-to-video jobs.

• Workers can now serve personal LoRAs that customers upload and Sogni approves. Files that do not match the model are refused cleanly and the job is retried elsewhere.
• LTX audio-to-video now animates to the uploaded soundtrack.
• FlashVSR video upscaling accepts detail, speed and seed options. Requests that use them go only to 1.0.212 or newer.
• MiniMax H3 gains a 2K output option, rolling out on FastH3 Turbo first. Peak GPU memory stays within the base job's.
• Nosana MiniMax deployments with 31 GB or more of RAM also serve FlashVSR. Pre-mount the pack to avoid a 6.4 GB download on restart.
• A missing license file no longer blocks a workflow.
• Disk and settings: new setting Personal LoRA cache (GiB) (PERSONAL_LORA_CACHE_GB), default 25 GiB, adjustable from the Dashboard. Set it to 0 to opt out of personal LoRA jobs. MiniMax H3 workers download a 690 MB upscaler file on their next model scan.

#Version 1.0.211

Released: September 10, 2026 PT · Update: Recommended for all workers. Required for FlashVSR video upscaling.

• New job type: FlashVSR upscales finished videos to 1080p or 1440p. Served by default on NVIDIA cards with at least 23 GB of VRAM and an Ampere or newer GPU. Other workers skip it and its download.
• Disk and settings: eligible workers download about 6.4 GB on their next model scan. No new settings.

#Version 1.0.210

Released: September 8, 2026 PT · Update: Recommended for all workers. Required for standalone BiRefNet background removal.

• Background removal no longer fails on simple subjects. Those failures were also taking healthy workers out of background-removal work for an hour at a time.
• Pixal3D keeps only its prompt-free reconstruction path.
• Disk and settings: nothing new.

#Version 1.0.209

Released: September 8, 2026 PT · Update: Recommended for all workers. Required for Qwen3-TTS, standalone BiRefNet background removal, or Pixal3D. Includes the 1.0.208 MiniMax H3 fix, so hosted deployments should move directly here.

• Qwen3-TTS adds three opt-in speech workflows on CUDA 13 workers with at least 24 GB of VRAM.
• Standalone BiRefNet background removal now delivers its transparent PNG correctly.
• Pixal3D handles portrait inputs correctly. It remains CUDA 13 only, local ComfyUI only, 30 GB of VRAM or more.
• Disk and settings: nothing downloads unless you select the workflow. Qwen3-TTS is about 12.7 GB, BiRefNet about 450 MB, Pixal3D about 13.4 GB.

#Version 1.0.208

Released: September 8, 2026 PT · Update: Superseded by 1.0.209, which carries the same fix. Update directly to 1.0.209.

• MiniMax H3 LightX2V Turbo returns to its earlier acceleration LoRA, which gives better motion and fewer artifacts. LightX2V Turbo jobs go only to 1.0.208 or newer.
• Pixal3D no longer needs a text prompt.
• Disk and settings: MiniMax H3 workers swap one 1.8 GB file for another of the same size. No new settings.

#Version 1.0.207

Released: September 8, 2026 PT (beta) · Update: Evaluation only. latest stayed on 1.0.206. Pin 1.0.207 or the beta tag to try it.

• New opt-in workflow: Pixal3D turns one image into a textured 3D model. Offered only to CUDA 13 workers with at least 30 GB of VRAM whose ComfyUI runs on the same machine. Select it under Workflows this worker serves to download about 13 GB.
• A worker opted into a pack with duplicate file names no longer runs out of work and restarts.
• Z-Image Turbo accepts up to 12 steps.
• SAM 3 image segmentation no longer fails on some results.
• Disk and settings: nothing new unless you opt into Pixal3D.

#Version 1.0.206

Released: September 7, 2026 PT · Update: Recommended for all workers, especially MiniMax H3 workers. Pinned deployments should move to 1.0.206.

• MiniMax H3 Turbo returns to its earlier attention settings after reports of artifacts. Longer clips return to their earlier render time.
• Interrupted model downloads recover more reliably.
• Disk and settings: nothing new.

#Version 1.0.205

Released: September 6, 2026 PT · Update: Recommended for all workers. Pinned deployments should move directly to 1.0.205; do not pin 1.0.204.

• FastH3 renders with one user LoRA are faster on 32 GB and larger cards. One sample went from 160 to 106 seconds.
• New setting Excluded LoRAs (EXCLUDED_LORA_IDS) keeps specific LoRAs from being assigned or downloaded, and removes a file only they used once the worker is idle.
• Opt-in workflow packs are selected with the existing Workflows this worker serves setting.
• The Windows installer reports the RAM actually available to Docker and can update .wslconfig for you when run interactively. CUDA 13 selection now checks every GPU's compute capability.
• Disk and settings: nothing downloads because of this update.

#Version 1.0.204

Released: September 6, 2026 PT · Update: Superseded by 1.0.205 before it was recommended. Do not pin it.

• Published during an interrupted release. Its changes ship in 1.0.205.

#Version 1.0.203

Released: September 5, 2026 PT · Update: Recommended for all workers. Important for 1.0.193 and 1.0.194, which can no longer serve Dark Beast Krea 2. Carries everything in 1.0.201 and 1.0.202.

• MiniMax H3 Balanced: only reference-to-video keeps the community adapter. The other Balanced modes return to the LightX2V LoRA and to their 1.0.199 speed.
• Workers stop offering models whose files were retired from the catalog, and report a clear error if a file is missing, instead of failing job after job. Sogni stops sending that model to the worker.
• Corrupt model files repair themselves.
• Long MiniMax H3 Turbo clips render about 17% faster with less flicker.
• Disk and settings: re-downloads a 1.8 GB MiniMax H3 LoRA if a previous release removed it. No new settings.

#Version 1.0.202

Released: September 5, 2026 PT · Update: Superseded by 1.0.203 before it was recommended. Do not pin it.

• A startup check withheld the ACE-Step 1.5 workflows. Fixed in 1.0.203, which carries the rest of its changes.

#Version 1.0.201

Released: September 5, 2026 PT · Update: Superseded by 1.0.203 before it was recommended. Do not pin it.

• Made long MiniMax H3 Turbo clips faster and steadier. Ships in 1.0.203.

#Version 1.0.200

Released: September 4, 2026 PT · Update: Recommended for MiniMax H3 workers, and needed to keep serving H3 Balanced.

• Every MiniMax H3 Balanced mode moves to the community adapter artists preferred. Balanced renders take about a third longer at the default size.
• Disk and settings: removes two 1.8 GB LoRA files and downloads a 744 MB adapter on the next start. No new settings.

#Version 1.0.199

Released: September 4, 2026 PT · Update: Recommended for all workers. Important on RTX 4090 and RTX 5080 hosts.

• Model switches are reported with their real duration. Sogni now sends cold jobs to the workers that switch fastest and asks slow switchers to switch less often.
• RTX 4090 workers get 2 GB more VRAM for models. RTX 5080 workers get their own memory profile.
• Disk and settings: nothing new.

#Version 1.0.198

Released: September 4, 2026 PT · Update: Recommended for all workers. Apple Silicon hosts should take it.

• Image renders finish about two seconds faster when the artist has the Sensitive Content Filter off. What gets withheld is unchanged.
• Mac workers: quantized models no longer crawl. NVIDIA hosts are unaffected.
• Disk and settings: nothing new.

#Version 1.0.197

Released: September 4, 2026 PT · Update: Recommended for all workers. Important on 24 GB and larger cards.

• A worker no longer drops its connection, and the job it was holding, while loading a large model.
• Workers now stay on one model family for longer, so the faster LoRA renders from 1.0.196 stick.
• Chroma image-to-image works again.
• Mac workers no longer accept extra ComfyUI start-up arguments from the environment.
• Disk and settings: nothing new.

#Version 1.0.196

Released: September 4, 2026 PT · Update: Recommended for all workers. Worth taking promptly on 24 GB cards.

• LoRA renders on quantized models are much faster on 24 GB cards. On an RTX 4090 one test went from 36 seconds to under 9.
• RTX 3090 and RTX 3080 workers stop advertising Z-Image and Qwen Image Edit, which returned corrupt output on those cards. Everything else is unchanged and the worker is not faulty.
• A CUDA 13 worker on a driver older than R580 now holds with a clear log line instead of restarting in a loop. Use the CUDA 12 image on those hosts.
• Disk and settings: nothing new.

#Version 1.0.195

Released: September 3, 2026 PT · Update: Recommended for all workers. latest moved here directly from 1.0.193.

• Dark Beast Krea 2 moves to the 3.0 checkpoint, which fixes blown-out skin and waxy textures. Renders are slower on the CUDA 12 track than on CUDA 13.
• MiniMax H3 reference-to-video Balanced is faster at every clip length.
• LTX-2 video-to-video no longer exhausts host RAM on long uploads.
• Time estimates after a model switch are accurate.
• Disk and settings: new setting EXCLUDED_WORKFLOW_IDS excludes specific models without giving up automatic model choice, and frees the disk only they used. Downloads a 13.2 GB Dark Beast file and removes the old one.

#Version 1.0.194

Released: September 3, 2026 PT · Update: Recommended for hosted deployments. latest did not move to it.

• A worker now checks at startup whether its machine is big enough and stops if not, so the provider can move it somewhere better.
• Custom LoRAs on every FastH3 mode now run on 24 GB CUDA 13 cards, including the RTX 4090 and RTX 3090.
• Disk and settings: optional MIN_VIABLE_DISK_GB, MIN_VIABLE_RAM_GB and HEALTH_PROBE_PORT for hosted deployments. Nothing changes without them.

#Version 1.0.193

Released: September 1, 2026 PT · Update: Required for MiniMax H3 Turbo and FastH3 Turbo workers. Recommended for all.

• Fixes the visual artifacts from 1.0.191 and 1.0.192.
• FastH3 Turbo is a new option about twice as fast as LightX2V Turbo. It runs on 24 GB CUDA 13 cards, including the RTX 4090 and RTX 3090. Custom LoRAs need 32 GB. Not offered on the CUDA 12 track.
• Disk and settings: nothing new.

#Version 1.0.192

Released: September 1, 2026 PT · Update: Recommended for all workers. Required for new MiniMax H3 Turbo workers.

• Large model downloads resume after a network interruption instead of starting over, so fresh workers enter service reliably.

#Version 1.0.191

Released: September 1, 2026 PT · Update: Superseded by 1.0.193. Do not use it for MiniMax H3 Turbo.

• Its faster MiniMax H3 Turbo path produced visual artifacts. 1.0.193 corrects it.

#Version 1.0.190

Released: August 31, 2026 PT · Update: Required for new or ephemeral workers on 1.0.189. Recommended for all.

• Fixes a startup regression that could leave a new worker idle instead of downloading its models and taking jobs.
• Restores installer-managed upgrades on Windows startup, and fixes live Dashboard settings updates.

#Version 1.0.189

Released: August 31, 2026 PT · Update: Recommended for all workers.

• Hardware care: the worker pauses new jobs when the GPU reaches about 189°F and resumes near 174°F, and takes a one-minute rest after 30 minutes of actual rendering. It never interrupts an active job. Both limits are adjustable from the Dashboard, and a natural idle gap already counts as the rest.
• Supported worker settings can be managed from the Dashboard.
• Disk and settings: the render-time rest and thermal settings are new. No downloads.

#Version 1.0.188

Released: August 30, 2026 PT · Update: Superseded by 1.0.189. If you are on 1.0.186 or 1.0.187, update.

• Turns off an over-cautious check that made capable GPUs decline jobs they could complete.

#Version 1.0.187

Released: August 30, 2026 PT · Update: Superseded. Update to the current release.

• First correction for workers declining valid jobs. Still too cautious; fixed in 1.0.188.

#Version 1.0.186

Released: August 29, 2026 PT · Update: Superseded. Update to the current release.

• Better memory handling and recovery, especially on Windows, but one new check declined some valid jobs. Corrected in 1.0.187 and 1.0.188.

#Version 1.0.185

Released: August 28, 2026 PT · Update: Superseded by 1.0.189.

• Refreshed the MiniMax H3 Balanced acceleration files. Replaced files were cleaned up automatically.

#Version 1.0.184

Released: August 28, 2026 PT · Update: Superseded by 1.0.189.

• Broad reliability update for MiniMax H3: previews, recovery, monitoring and compatibility.

#Version 1.0.183

Released: August 27, 2026 PT · Update: Superseded by 1.0.189.

• Fixed a MiniMax H3 failure after a model was unloaded and reloaded, present in 1.0.181 and 1.0.182.

#Older releases

Every version below is superseded. A worker updates straight to the current release; it never needs an intermediate version.

1.0.151 and 1.0.153 have no public image.