Video GenerationLightricksother

Lightricks/LTX-2.5

Lightricks/LTX-2.5

Parameters

22B

Context

Downloads

39

Likes

349

Architecture

View on HuggingFace

GPU Requirements

VRAM / RAM / disk estimates by quantization and usage scenario

QuantizationMinimumRecommendedProduction
FP1616bit
86.6 GB
RAM 130.0 · Disk 102.9
94.0 GB
RAM 141.0 · Disk 128.9
105.3 GB
RAM 158.0 · Disk 167.9
FP88bit
46.1 GB
RAM 69.1 · Disk 71.0
51.6 GB
RAM 77.4 · Disk 97.0
61.1 GB
RAM 91.6 · Disk 136.0
INT88bit
46.1 GB
RAM 69.1 · Disk 71.0
51.6 GB
RAM 77.4 · Disk 97.0
61.1 GB
RAM 91.6 · Disk 136.0
AWQ4bit
26.5 GB
RAM 39.8 · Disk 55.6
31.2 GB
RAM 46.8 · Disk 81.6
39.8 GB
RAM 59.6 · Disk 120.6
GPTQ4bit
26.8 GB
RAM 40.2 · Disk 55.8
31.5 GB
RAM 47.2 · Disk 81.8
40.0 GB
RAM 60.1 · Disk 120.8
GGUF4bit
27.3 GB
RAM 41.0 · Disk 56.2
32.0 GB
RAM 48.0 · Disk 82.2
40.6 GB
RAM 60.9 · Disk 121.2

GPU Recommendations

Best ValueBEST

1× H100 SXM

80 GB VRAM

$1,170/mo

≈ 288 video-sec/GPU-h

Effective production throughput ≈ 172.8 video-sec/GPU-h

score 0.5935

Deploy on vast ↗
Cheapest

1× RTX A6000

48 GB VRAM

$226/mo

≈ 288 video-sec/GPU-h

Effective production throughput ≈ 172.8 video-sec/GPU-h

score 0.6641

Deploy on vast ↗
Performance

8× H200 SXM

1128 GB VRAM

$20,440/mo

≈ 288 video-sec/GPU-h

Effective production throughput ≈ 172.8 video-sec/GPU-h

score 0.2197

Min Complexity

1× RTX 5090

32 GB VRAM

$257/mo

≈ 288 video-sec/GPU-h

Effective production throughput ≈ 172.8 video-sec/GPU-h

score 0.6441

Deploy on vast ↗

💬 Community — real deployment experience

0 Articles · 0 Benchmarks

View Community →

GPUs that fit (single card)

RTX A6000

48 GB VRAM

$226

/mo · vast

RTX 5090

32 GB VRAM

$257

/mo · vast

L40S

48 GB VRAM

$341

/mo · vast

RTX PRO 6000 WS

48 GB VRAM

$482

/mo · vast

A100 80GB

80 GB VRAM

$876

/mo · vast

AmciHub

AI Model Deployment & Compute Intelligence — analyze AI model requirements, GPU performance, cloud pricing and deployment costs to find the right deployment solution.

© 2026 AmciHub. All rights reserved. Model → Requirement → GPU → Benchmark → Cloud → Cost → Recommendation