Runway Brings Gen-4.5 and Aleph 2.0 to Private Enterprise Deployments
Runway is opening up its Gen-4.5 and Aleph 2.0 video models for private enterprise deployment on Baseten with fine-tuning and reserved GPUs.
- Runway partners with Baseten to offer Gen-4.5 and Aleph 2.0 as dedicated deployments.
- Models run in single-tenant VPCs with reserved GPU capacity and autoscaling.
- Enterprises can fine-tune weights on proprietary footage under a direct Runway license.
- Aleph 2.0 handles multi-scene edits up to 30 seconds at 1080p with cross-shot consistency.
- Infrastructure covered by SOC 2 Type II, HIPAA, and a 99.99% uptime SLA.
- Existing Runway enterprise users include Amazon, Microsoft, Robinhood, and Dolce & Gabbana.
Runway has primarily distributed its generative video models through its hosted platform. A partnership with Baseten now gives enterprises dedicated, single-tenant deployments of Gen-4.5 and Aleph 2.0, including access to the model weights, fine-tuning on proprietary footage, and reserved GPU capacity.
The Baseten partnership is Runway’s first dedicated deployment option outside its own infrastructure. It gives studios, brands, and regulated companies more control over workload isolation, training data, and production capacity.
From license to endpoint
- License: The customer licenses the model weights directly from Runway.
- Deploy: Baseten runs the models in a secure, single-tenant environment with dedicated GPUs.
- Customize: The customer can fine-tune the models on proprietary footage and serve the resulting checkpoints inside the same tenant.
- Operate: Runway and Baseten jointly support customization, deployment, scaling, and production inference.
Generation and editing split the work
| Model | Inputs | Primary role | Production details |
|---|---|---|---|
| Gen-4.5 | Text and images | Generate new video | Supports scene direction, camera movement, and varied visual styles |
| Aleph 2.0 | Source video, prompts, or edited keyframes | Edit and restyle existing video | Supports multi-scene video up to 30 seconds at 1080p |
Gen-4.5 can animate product shots, turn storyboards into video, and produce campaign variants. Aleph 2.0 can replace an object, alter a scene’s appearance, or propagate a keyframe edit across relevant shots while preserving surrounding details.
Aleph’s cross-shot consistency addresses a persistent production problem: an edited subject must retain its identity as camera angles, lighting, and backgrounds change. That capability makes the model useful for sequences that extend beyond isolated clips.
Dedicated GPUs keep workloads separate
Shared inference services can expose studios and brand teams to unacceptable controls around unreleased intellectual property, licensed talent likenesses, and pre-launch product footage. Baseten assigns each customer an isolated workload plane with dedicated GPU nodes while operating the surrounding control plane, which manages deployment, scheduling, scaling, and monitoring.
Baseten governs access to Runway’s weights through a distribution agreement. The deployments run on infrastructure covered by Baseten’s SOC 2 Type II and HIPAA compliance programs, and Baseten provides a 99.99% uptime service-level agreement. Customers remain responsible for configuring their applications and data practices to meet their own regulatory obligations.
Fine-tuning stays inside the tenant
Fine-tuning adjusts a model with customer-provided examples so its output better follows a company’s visual language, product catalog, characters, or content rules. A retailer could train Gen-4.5 toward consistent product presentation, while a studio could adapt Aleph 2.0 to preserve specific objects and characters during edits.
The customer’s direct Runway license governs the fine-tune. Baseten supplies the infrastructure for training and inference, while the resulting checkpoints and derivative weights remain inside the customer’s tenant.
Runway removes an enterprise constraint
Amazon, Robinhood, Dolce & Gabbana, and Microsoft are among the enterprise creative teams using Runway in production. Dedicated deployment gives security and procurement teams a path to approve workloads involving sensitive media, proprietary training sets, and fixed capacity requirements.
Baseten, previously associated mainly with language-model and audio inference, gains a demanding video workload. Video generation consumes substantial GPU memory and compute, making reserved capacity and autoscaling central to latency, throughput, and cost planning.
What builders gain
- Isolated inference: Applications can call Runway models within a dedicated workload environment instead of routing media through a shared service.
- Private customization: Teams can fine-tune models on proprietary footage and retain the resulting weights inside their tenant.
- Predictable capacity: Reserved GPUs provide a production baseline, with autoscaling available for demand spikes.
- Enterprise controls: SOC 2 Type II infrastructure, HIPAA-aligned operations, and a 99.99% uptime SLA support security and procurement reviews.
- Joint support: Runway handles model licensing and expertise, while Baseten manages deployment and inference infrastructure.
Runway Dev remains the simpler route for teams that can use a hosted API. The Baseten deployment targets organizations that require isolated infrastructure, custom weights, reserved compute, and direct control over sensitive video workloads.
Comments
