llms.help
Search/
RU

Model profile

Gen-4.5

Built by Runway

Gen-4.5 is Runway’s video-generation model for text-to-video and image-to-video workflows. Runway’s API model directory lists the identifier gen4.5 with text or image input and video output.[1]

ReleasedDec 1, 2025[1]
CategoryVideo Generation
Directory statusCatalogued
Official model cardVisit source ↗

Developer snapshot

Key specifications

Verified Jul 18, 2026

Modality
Text/image to video[1]
API model ID
gen4.5[1]
Released
December 1, 2025[1]
Output
Video[1]
Access
Runway product and API[1]
Known limits
Causality and object permanence[1]

Technical profile

What developers should know

01

Current API identity

Runway’s current model directory lists gen4.5 under video generation and accepts either text or image input. The same directory separates other video, image, upscale, audio, and real-time models rather than treating Gen-4.5 as a universal media endpoint.[1]

02

Provider-reported strengths

Runway’s launch material emphasizes motion quality, prompt adherence, visual fidelity, temporal consistency, physical behavior, character expression, and stylistic control. These descriptions are vendor-reported capabilities and should be validated on the intended shot types.[2]

03

Disclosed failure modes

Runway explicitly identifies causal-order mistakes, failures of object permanence across frames, and a bias toward actions succeeding as limitations. Evaluation clips should therefore include occlusion, multi-step cause and effect, and failed-action scenarios.[2]

Choose the artifact

Official checkpoints

Use the exact documented model or endpoint identifier and record input assets, parameters, output metadata, and failure states for reproducible media workflows.[1] [2]

Runway video API gen4.5

Text-to-video and image-to-video generation through Runway’s task API.[1]

Technical launch note Gen-4.5

Provider examples, capability discussion, deployment notes, and disclosed limitations.[2]

Deployment note

Test temporal consistency and causal behavior on representative shots before adopting the model in a production workflow. Save the exact API model ID and input assets with each evaluation.[1] [2]