Skip to content
sekushi
18+Join freeMy unlocks

H3 model file names decoded: bf16, fp8, int8, convrot, nvfp4

A beginner's guide to MiniMax H3's file names, with pictures: what pruned, bf16, int8_convrot, fp8_scaled, w6a8 and nvfp4 mean, and which suit your card.

Published · 8 min read

ComfyUI's MiniMax H3 download page lists a dozen video model files whose names differ only at the end. Most of them are the same model stored in different ways.

This guide takes a file name apart piece by piece, says what each way of storing costs, and ends with the files to download for 16, 24 and 32 GB cards. If you are new to H3, read MiniMax H3 in ComfyUI: what every part does first.

A few words first

  • Model file: a large file of learned numbers, called weights. H3 needs four: the video model, the text encoder and two decoders.
  • Bit: the smallest unit of computer memory. A number stored in 16 bits is more exact than one stored in 8 or 4, and takes more room.
  • Precision: how many bits each weight is stored in, and in what format.
  • Floating point: a way of storing numbers with a movable decimal point, such as 0.0031 or 2,400. Integer formats store whole numbers only.
  • Quantised: stored in fewer bits than the original, to make the file smaller.
  • VRAM: the memory on your graphics card. RAM is your computer's main memory.
  • GB: gigabyte. File listings count a thousand million bytes; cards count 1,073,741,824 bytes (a binary gigabyte), so a 24 GB card holds about 25.8 GB of files.

Reading a file name

A file name is five pieces, read left to right. Here is the default video model in ComfyUI's templates.

A MiniMax H3 file name, piece by pieceminimax_h3_fl2va_pruned_int8_convrot.safetensors reads as five pieces. minimax_h3 is the model family. fl2va is the checkpoint, fl2va for text and frames or ref2va for references. pruned means precomputable weights were left out. int8_convrot is how each number is stored: 8-bit integers, rotated first. .safetensors is the file format, which holds numbers only and no code.minimax_h3_fl2va_pruned_int8_convrot.safetensors1minimax_h31. the model familyMiniMax H32fl2va2. which checkpointfl2va (text, frames) or ref2va3pruned3. precomputable weights left outmissing from the full files4int8_convrot4. how each number is stored8-bit integers, rotated first5.safetensors5. the file formatnumbers only, no code inside
The default video model's name, taken apart. Text encoder and decoder names follow the same pattern.
  1. minimax_h3 is the model family.
  2. fl2va says which of the two model files this is: fl2va for text and frames, or ref2va for references. H3's modes explained covers the difference.
  3. pruned means a block of weights that can be worked out in advance was left out. The full files do not have this word.
  4. int8_convrot is the precision: 8-bit integers, with a rotation trick explained below.
  5. .safetensors is the file format. It holds numbers only, no code, so it is safe to load.

The text encoder's name works the same way: qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors is the Qwen3-VL-32B encoder, cut to the layers H3 uses, stored in 4-bit NVFP4 with AWQ.

Where the names appear in ComfyUI

The template's main box has one row for each model file. The file you choose there must be in the matching folder inside ComfyUI's models folder.

Four rows of the main H3 box, numbered 1 to 4, each showing a model file name
The file rows of ComfyUI's stock Image to Video template, with its default files.
  1. unet_name is the video model. Its file goes in diffusion_models.
  2. clip_name is the text encoder. Its file goes in text_encoders.
  3. vae_name is the video decoder. Its file goes in vae.
  4. audio_vae is the audio decoder. Its file also goes in vae.

Click a row to pick a different file: the list shows the files in that folder.

What every precision word means

Fewer bits make a smaller file that needs less memory, at some cost in accuracy. The question is always how much.

In the nameWhat it isBits per weightNotes
bf16bfloat16, a 16-bit floating-point format16MiniMax's own release precision; every other file is compared with it
fp16half precision, 16-bit floating point16the video decoder, at the precision MiniMax's own code runs it
fp32single precision32the audio decoder; small enough not to matter
int8_convrot8-bit integers, after a rotation8ComfyUI's recommended format for the video model
fp8_scaled8-bit floating point, with scale factors8the fallback for setups that cannot run int8_convrot
w6a86-bit weights, 8-bit working numbers6for smaller cards; expect some loss
nvfp4NVIDIA's 4-bit floating-point format4native on RTX 50-series (Blackwell) cards
awqactivation-aware weight quantisationn/aa method for choosing which weights to protect at 4 bits, not a format

int8_convrot, in plain words

Squeezing weights into 8-bit integers works badly when a few values are far larger than the rest: one big number sets the scale, and everything else is rounded coarsely. ConvRot first rotates groups of weights (community repackagers describe a group-wise Hadamard rotation), which spreads those large values out, then rounds. While the model runs, the matching rotation is applied on the way in, and the card's fast 8-bit units do the multiplying.

ComfyUI's card for its H3 files says to prefer int8_convrot for the video model if your PyTorch is built for CUDA 13.0 (cu130), and to use fp8_scaled only if it is not. A published community comparison points the same way: on one seed of a text-to-video clip, the 8-bit ConvRot output stayed close to bf16, while fp8 drifted further. One seed is a small test, but it agrees with ComfyUI's advice.

pruned, in plain words

H3 has a large set of weights in the branches that tell it how far along the denoising is (called AdaLN). MiniMax's card puts them at about 13 billion of the model's 33 billion numbers, and says their outputs can be worked out in advance. ComfyUI's "pruned" files replace those weights with a short table of precomputed curves.

A published community measurement found the table's error far smaller than the rounding that 16-bit storage already causes. On that evidence, the unpruned files are not an upgrade; they are only bigger.

One caveat: a LoRA (a small add-on file) trained on the unpruned model may adjust those AdaLN weights. On a pruned file that part has nothing to attach to, and ComfyUI skips it with a shape-mismatch warning.

Every official file, with its size

The chart sets the video model and text encoder files against the memory of three common cards.

MiniMax H3 file sizes against card memoryVideo model files: bf16 66.3 GB, pruned bf16 40.2, int8 convrot 34.0, pruned int8 convrot 21.0 (the default), pruned fp8 scaled 21.0, pruned w6a8 16.0. Text encoder files: bf16 51.5, int8 convrot 27.1, nvfp4 awq 15.7 (the default). A 16 GB card holds 17.2 decimal gigabytes, a 24 GB card 25.8 and a 32 GB card 34.4. Only the 6-bit video file and the 4-bit encoder are smaller than a 16 GB card; the default video model is smaller than a 24 GB card.file size in GB; the lines are each card's memory162432Video model (each ref2va twin is the same size)bf1666.3pruned_bf1640.2int8_convrot34.0pruned_int8_convrot21.0pruned_fp8_scaled21.0pruned_w6a816.0Text encoder (Qwen3-VL-32B)bf1651.5int8_convrot27.1nvfp4_awq15.7rose: the template's default file
Official file sizes from ComfyUI's download page, in decimal GB. The dashed lines are the memory of 16, 24 and 32 GB cards.

A file longer than your card's line does not fit on the card whole. That is normal: ComfyUI keeps the rest in RAM and streams it in while it works. Running H3 on 16, 24 and 32 GB cards explains how.

All files are from ComfyUI's repository (Comfy-Org/MiniMax-H3), sizes from its Hugging Face listing. Each fl2va file has a ref2va twin of the same size.

Video model (folder diffusion_models)

FileSize
minimax_h3_fl2va_bf1666.28 GB
minimax_h3_fl2va_pruned_bf1640.23 GB
minimax_h3_fl2va_int8_convrot34.04 GB
minimax_h3_fl2va_pruned_int8_convrot (templates' default)20.97 GB
minimax_h3_fl2va_pruned_fp8_scaled20.96 GB
minimax_h3_fl2va_pruned_w6a815.98 GB

Text encoder (folder text_encoders): qwen3vl_32b_minimax_h3_bf16 51.51 GB, …_int8_convrot 27.14 GB, and …_nvfp4_awq 15.69 GB (templates' default).

Decoders (folder vae): minimax_h3_video_vae_fp16 5.21 GB, minimax_h3_video_vae_int8_convrot 2.81 GB (templates' default) and minimax_h3_audio_vae_fp32 0.61 GB.

Speed-up LoRAs (folder loras), about 1.96 GB each: minimax_h3_fl2v_turbo_8step_v1.0, minimax_h3_fl2v_turbo_4step_v1.0_768p and minimax_h3_ref2v_turbo_4step_v0.1, all from the lightx2v team.

The text encoder: does 4-bit hurt?

The 4-bit encoder is the most squeezed file in the default set, and it reads both your prompt and your start picture. The evidence so far:

  • ComfyUI ships it as the default and says it does not need a Blackwell card.
  • One quantiser compared a 4-bit build (a slightly different one) with bf16 on a set of clip pairs and reported them alike to the eye.
  • One tester reported that 4-bit loses coherence on tricky setups, without examples.
  • No maker has published a controlled comparison on long prompts.

The encoder runs once per prompt, not once per step, so a larger encoder costs memory more than time. If you have the RAM, render the same prompt and seed with the 8-bit file and compare.

Files from outside ComfyUI's repository

  • NVFP4 video models (about 12.5 GB): fast on Blackwell cards, slow on older ones. One such build's own card says 4-bit weights appear to cost some motion quality.
  • GGUF files (Q8_0 down to Q2_K, about 21 to 7 GB), loaded through a custom node. They expand their weights at every step, which tends to make them slower, and the comparison above measured an 8-bit weight-only file as further from bf16 than int8_convrot.
  • Merges and fine-tunes mix other weights into H3. Judge each by its own page and your own tests.

Which files to download

The video model sets your quality ceiling; the other files are mostly about memory.

Your cardVideo modelText encoderVideo decoder
32 GB (RTX 5090)pruned_int8_convrotnvfp4_awqfp16
32 GB, 128 GB of RAMpruned_bf16nvfp4_awqfp16
24 GB (RTX 4090, 3090)pruned_int8_convrotnvfp4_awqfp16
16 GBpruned_int8_convrotnvfp4_awqfp16
16 GB, short of RAMpruned_w6a8nvfp4_awqint8_convrot
No cu130 PyTorchpruned_fp8_scalednvfp4_awqfp16

Why these:

  • 32 GB: the 8-bit file is ComfyUI's recommended format. With plenty of RAM, the 16-bit file is MiniMax's own precision; it is slower, and the measured gain is small.
  • 24 GB and 16 GB: the default set works on both if you have enough RAM; ComfyUI streams the part of the video model the card cannot hold. Short of RAM, the 6-bit file and the smaller decoder lighten the load, with some loss that ComfyUI does not put a number on.
  • System RAM matters more than VRAM. The files in the 24 GB row add up to about 42.5 GB. If your RAM cannot hold them alongside everything else, files are read from disk again and renders slow down badly.
  • A smaller file is not a faster file while everything fits in RAM. ComfyUI's H3 implementer has explained that streaming is hidden behind the computing, so step time is set by the arithmetic, not the file size.
  • The 8-bit video decoder saves memory, not quality. Reports on its look are mixed; use fp16 unless memory forces the smaller one.
  • Keep the pair matched: fl2va files with fl2v_… LoRAs, ref2va files with ref2v_… LoRAs.

Sources: the file listing and card of Comfy-Org/MiniMax-H3; MiniMax's H3 model card; ComfyUI's H3 tutorial; the pruning and precision measurements at multimodalart/MiniMax-H3-Pruned; the 4-bit video build's card at lilcheaty/MiniMax-H3-NVFP4; GGUF files at unsloth/MiniMax-H3-GGUF. Sizes as listed in October 2026. Screenshot: ComfyUI's stock "MiniMax H3: Image to Video" template (ComfyUI 0.39.2), labels added.