Mochi / VIDEO
Mochi 1 Preview VRAM requirements
Explore memory tradeoffs with sourced model data and transparent estimates.
FP8 estimate
144.5 GiB
at default settings
FP16 estimate
144.5 GiB
at default settings
Suggested GPU
1 × NVIDIA HGX B200 180 GiB
Model files & facts
| Media | video |
|---|---|
| Default resolution | 848 × 480 |
| Default frames | 163 |
| Temporal compression | 6 |
| Notes | Component-based diffusion memory estimate. |
| Component | Weight file (source) | Size |
|---|---|---|
| Backbone FP16 | split_files/diffusion_models/mochi_preview_bf16.safetensors | 18.68 GiB |
| Backbone FP8 | split_files/diffusion_models/mochi_preview_fp8_scaled.safetensors | 9.34 GiB |
| T5-XXL FP16 | split_files/text_encoders/t5xxl_fp16.safetensors | 9.12 GiB |
| T5-XXL FP8 | split_files/text_encoders/t5xxl_fp8_e4m3fn_scaled.safetensors | 4.80 GiB |
| VAE | split_files/vae/mochi_vae.safetensors | 0.86 GiB |
Precision & offloading comparison
Source default resolution and frame count · batch 1 · FP16 text encoders · safety margin included.
| Backbone | Text encoders | Total |
|---|---|---|
| FP16 | GPU | 144.51 GiB |
| FP16 | CPU | 144.51 GiB |
| FP8 | GPU | 144.51 GiB |
| FP8 | CPU | 144.51 GiB |
ComfyUI file placement
Example component locations; install matching loader nodes. Brace patterns in source filenames denote all shards. Preserve the source layout when a loader requires it.
models/diffusion_models/split_files/diffusion_models/mochi_preview_bf16.safetensors
models/diffusion_models/split_files/diffusion_models/mochi_preview_fp8_scaled.safetensors
models/text_encoders/split_files/text_encoders/t5xxl_fp16.safetensors
models/text_encoders/split_files/text_encoders/t5xxl_fp8_e4m3fn_scaled.safetensors
models/vae/split_files/vae/mochi_vae.safetensorsRecommended GPUs by memory fit
FP8 backbone at source default resolution, FP16 text encoders on GPU. Smallest nominal memory first; green results only. Check backend and precision support separately.
1 × NVIDIA HGX B200 180 GiB
Fits with safety headroom
1 × GB200 Blackwell GPU 186 GiB
Fits with safety headroom
1 × NVIDIA B100 (192 GiB)
Fits with safety headroom
Sources & verification
Exact file sources are linked in the component table above. ComfyUI memory planner constant
Last verified: 2026-11-20
Try your configuration
Your workload
Custom diffusion model parameters
Uses the ComfyUI activation and decode profile for Mochi 1 Preview. Leave fields empty to use catalog files. Placeholder counts approximate half-precision file sizes.
Your hardware
Output dimensions
Images or videos per batch. Serving concurrency and throughput require separate measurements.
Resolution and peak VRAM
Sampling and timing
One step is one scheduler iteration. Enter timings from the same device, batch, resolution and guidance settings. Fixed overhead includes encoding, decoding and other work outside sampling.
Advanced
Performance and memory results
Memory estimate
144.5 GiB needed
Choose more memory, a smaller model, or a lighter workload.
RTX 4090 24 GiB
Vertical line = usable hardware memory
- VAE0.86 GiB
- VAE decode activations127.37 GiB
- ComfyUI inference reserve0.80 GiB
- Safety margin15.48 GiB
Per-stage peak
ComfyUI moves idle models to system RAM between stages; the largest stage sets the requirement. Totals include the safety margin.
- Text encode11.11 GiB
Text encoders (FP16) 9.12 + ComfyUI inference reserve 0.80
Subtotal 9.92 + safety margin (12%) 1.19 ≈ 11.11 GiB
- Sampling19.15 GiB
Diffusion model (FP8) 9.34 + Sampling activations 6.96 + ComfyUI inference reserve 0.80
Subtotal 17.10 + safety margin (12%) 2.05 ≈ 19.15 GiB
- VAE decodePeak144.51 GiB
VAE 0.86 + VAE decode activations 127.37 + ComfyUI inference reserve 0.80
Subtotal 129.03 + safety margin (12%) 15.48 ≈ 144.51 GiB
Exceeds device capacity by 105.029 GiB before safety margin
Generation performance
- Denoising steps
- 50
- Conditioning branches per step
- 1
- Latent patch tokens per sample
- 44,520
- Estimated backbone work per step
- 2,061.919 TFLOP
- Step time at single-device BF16 peak
- 12,481.347 ms
DiT work estimates cover linear layers and visual self-attention. Enter measured step time and fixed overhead under Sampling and timing.
Compare GPU capacity
Frequently asked questions
How much VRAM does Mochi 1 Preview need?
Our default estimate at the source default resolution, batch 1, FP16 text encoders on GPU and FP8 backbone is 144.5 GiB including the safety margin. These are estimates, not measured guarantees.
Can I offload text encoders?
Yes. The CPU option removes text encoder weights from the GPU estimate, trading GPU memory for system RAM and transfer time. Each component still needs to fit on its assigned device.
Does a green result guarantee compatibility?
No. Green means the estimated memory including safety headroom fits. Backend support, precision support, kernels and runtime allocation can differ. Confirm compatibility and measure your actual workflow.
Where do these estimates come from?
Architecture and file sizes are traced to model sources. The precision table identifies measured GGUF rates, block storage and nominal bit widths. The source data for Mochi 1 Preview was last verified on 2026-11-20. Memory estimates use GiB.