ComfyUI Error Messages: What Each One Means

Search this page for your exact string — but read one line of your own log first. "Node threw an error during execution." is not an error message. It is a fixed label from ComfyUI's frontend locale file, printed above every runtime failure, and the real message is the Exception Message field in the ComfyUI Error Report underneath it. Once you have that string, the fix is decided by which of four stages produced it: validation (nothing ran yet), model loading, sampling, or the memory and quantisation layer.

Most ComfyUI errors are too small to deserve their own article, but there are a lot of them, and the wording is specific enough that you can identify the failing subsystem from the string alone. That is what this page is for: a lookup table you can Ctrl+F, one line per string, linking out to the deep pages where a cause genuinely needs a thousand words.

How every row here is sourced. Each string is either quoted from the ComfyUI source line that raises it, or from a numbered issue on the tracker where somebody reported it verbatim. Source lines were read from the repository at commit 924743a (23 August 2026), which reports __version__ = "0.33.0" in comfyui_version.py. Nothing on this page is a reconstruction from memory, and where a string comes from a bug report rather than the source, the row says so and links the issue so you can check whether it has been closed since.

Error text drifts between releases. ComfyUI rewords its messages fairly often, so if your string is close to a row here but not identical, you are almost certainly in the right place with a different version. The find-it-yourself section at the bottom shows how to locate the exact line in your own install in about ten seconds.

Read the Error Report Before the Table

When a node fails at runtime, ComfyUI shows a red banner and a structured block that begins # ComfyUI Error Report. The banner text is generic. The block is the diagnosis.

# ComfyUI Error Report
## Error Details
- **Node ID:** 433
- **Node Type:** CLIPTextEncode
- **Exception Type:** ValueError
- **Exception Message:** ValueError: not enough values to unpack (expected 4, got 1)

That is quoted from issue #15797, whose title is — inevitably — "Node threw an error during execution." The reporter pasted the banner as the problem. The actual problem is four lines lower, in a completely different node than the banner suggests to most readers.

The banner string lives in ComfyUI_frontend's English locale file, alongside its longer sibling, "This node threw an error during execution. Check its inputs or try a different configuration." Both are static UI copy. Neither is produced by the code that failed.

So the reading order is: Exception Type and Exception Message first, Node Type second, stack trace third. The tables below are keyed on that Exception Message.

Errors That Fire Before Anything Runs

These are validation failures. ComfyUI rejects the graph before executing a single node, which is why they appear instantly with no progress bar and no VRAM movement. Every one of them is raised in execution.py, and they arrive wrapped in a single umbrella message.

The practical shortcut for this whole table: if the failure is instant and your GPU never spun up, it is a validation error, and the fix is in the workflow or the folder layout rather than in your hardware or flags.

Errors From Loading a Model File

These fire after validation passes, while a loader node reads a file off disk. The distinguishing feature is that they name a path.

Two of these six are worded as though your file is broken when it is not, which is why "re-download the model" is such a common piece of bad advice in this category. Check the file size before you spend an hour on a 15GB download.

Errors Raised While Sampling

These appear after the progress bar starts moving. The Node Type field in the error report tells you which node stopped, but with sampling failures the cause is usually a mismatch introduced several nodes upstream.

If your sampler completes with no error at all and hands you a pure black frame, that is a different failure with its own diagnosis — see why ComfyUI generates black images.

Errors From Memory Management and Quantised Weights

This is the newest and fastest-moving group, because ComfyUI's DynamicVRAM streaming layer and its quantisation kernels are both under active development. Expect these strings in particular to change.

Note the shape of the first row: a disk-read failure that presents as an out-of-memory error. If you are chasing OOM on a machine that has obvious free VRAM, search your log upward for read_file_slice before you start lowering resolution.

Warnings People Paste as Errors

These are logging.warning calls, not exceptions. They print in yellow, scroll past during a normal successful load, and get pasted into bug reports as the cause of unrelated problems.

The distinction is worth internalising, because two of the three above are noise and the third is the only clue you will get for a failure that produces no error at all.

How Do You Find Where Your Own Error Comes From

If your string is not on this page, you can find its origin faster than you can search for it. From your ComfyUI root:

grep -rn "shapes cannot be multiplied" --include="*.py" .

Three outcomes, and each one is informative:

• A hit under comfy/ or in execution.py or nodes.py. It is a core message, and the surrounding lines usually tell you the condition that triggered it.
• A hit under custom_nodes/. A node pack raised it. Take it to that pack's tracker, not ComfyUI's.
• No hit at all. The message came from PyTorch, NumPy, safetensors or transformers. mat1 and mat2 shapes cannot be multiplied and Torch not compiled with CUDA enabled are both in this category — they are PyTorch strings that ComfyUI merely surfaces.

That third case explains why so much ComfyUI troubleshooting advice misses: people search the ComfyUI tracker for an error PyTorch raised, and the top results are twenty unrelated workflows that happened to hit the same shape mismatch.

Before filing anything, do the triage the issue template asks for and disable custom nodes. Every bug report quoted on this page opens with that checkbox, and it is the fastest way to split "ComfyUI is broken" from "one of my forty node packs is broken". If you are still assembling a working install, the complete ComfyUI guide covers the layout these errors keep referring to, and the general local-AI troubleshooting guide covers the driver and environment failures that sit underneath all of it.

Which ComfyUI Version Do These Strings Come From

Source line numbers and message text on this page were read from the ComfyUI repository at commit 924743a, dated 23 August 2026, with comfyui_version.py reporting 0.33.0. The permalinks in every table are pinned to that commit, so they will keep pointing at the right line even after the file changes.

Three caveats that will save you time:

• Line numbers move constantly. Treat them as a starting point and grep for the string itself.
• Wording changes between releases. Several of the validation messages here were reworded during the v0.3x series. If yours differs slightly, the row still applies.
• Issue state changes. Issues linked above were open or closed as noted at the time of writing. Click through before you build a workaround into your workflow — several of these have already been fixed once.

Where a single string needed more than one sentence, it has its own page: IMPORT FAILED and missing nodes, out-of-memory after an update, black image output, and, for the quantisation rows, how GGUF, fp8 and Nunchaku builds differ.

Sources

• ComfyUI source at commit 924743a (v0.33.0, 23 August 2026): execution.py, nodes.py, comfy/sd.py, comfy/utils.py, comfy/ops.py, comfy/lora.py, comfy/memory_management.py
• ComfyUI_frontend English locale (the "Node threw an error during execution." banner text)
• ComfyUI docs — custom node troubleshooting (the disable-custom-nodes procedure every issue template links)
• Comfy-Org/ComfyUI issues #3161, #3377, #5508, #6414, #6476, #7177, #7433, #8438, #9318, #9423, #10266, #10891, #11452, #12132, #12648, #13954, #14250, #14806, #14829, #15118, #15255, #15400, #15433, #15599, #15797