What actually changed
The models are not new. Gen-4.5 has been generating video from text and image prompts inside Runway's own products, and Aleph 2.0 has been editing existing footage the same way, carrying a change made to one keyframe across the shots around it, up to 30 seconds at 1080p. What changed on October 14 is the address. Both models can now run inside a customer's own isolated corner of Baseten, instead of only inside Runway Creative or through the Runway Dev API.
Gen-4.5 and Aleph 2.0 are now offered through Baseten Dedicated Inference. Customers license the proprietary model weights directly from Runway, while Baseten runs an isolated workload plane with dedicated GPUs and manages the control plane. Baseten's access to the weights is governed by a distribution agreement, and both companies support deployment and customization. The isolation separates customer compute and scheduling from other tenants; it does not, by itself, establish every data-handling, retention or access restriction in a customer's eventual contract.
Why a studio would pay for a private room
The shared API is fine until the footage is not. Unreleased campaign material, licensed characters, and footage under embargo are the kinds of assets legal departments dislike sending to a multi-tenant endpoint, however good the vendor's paperwork is. A single-tenant deployment answers that objection in architecture rather than in a contract clause.
Dedicated inference may appeal to enterprises handling unreleased footage, licensed assets or workloads requiring isolated compute. Baseten says customers can fine-tune Runway's models on their own data under a direct Runway licence, with reserved GPU capacity and autoscaling. It also cites SOC 2 Type II and HIPAA compliance programs and advertises a 99.99% uptime SLA. Buyers should verify data access, retention, fine-tuning rights, capacity guarantees, SLA exclusions and remedies in the actual agreements; isolation and compliance programs do not automatically establish that every workload meets its regulatory obligations.
The part with no number on it
Baseten has not published a price, minimum spend or capacity commitment for this dedicated Runway offering. By comparison, Runway Dev publicly lists Gen-4.5 at $0.12 per generated second and Aleph 2.0 at $0.28 per second. Those usage-based API prices are not directly comparable with a dedicated deployment, which may involve reserved GPUs, licensing, support and contractual capacity terms. Baseten names Amazon, Robinhood, Dolce & Gabbana and Microsoft as existing Runway production users, but provides no evidence that any has adopted the newly announced dedicated service.
The model library lists both Runway models for deployment inquiries rather than checkout. Anyone pricing this route against the public API will have to run a sales process to get the other side of the comparison.
What to do with it
If isolated infrastructure, fine-tuning or reserved capacity is a requirement, request a dedicated-deployment proposal covering GPU allocation, model licensing, data handling, SLA remedies and minimum commitments. If predictable usage-based pricing matters more, start with Runway Dev's published API rates. Compare the total cost per usable finished video, including retries, idle reserved capacity and support, rather than only the price per generated second.
Sources
- [1] Baseten — "Deploy Runway's video models on Baseten", 14 October 2026 (primary announcement: dedicated single-tenant deployment, weights licensed directly from Runway, distribution agreement, SOC 2 Type II and HIPAA programs, fine-tuning, reserved capacity with autoscaling, 99.99% uptime SLA, named enterprise production users)Read source
- [2] Baseten model library — Gen-4.5 (dedicated deployment listing, offered for deployment inquiries)Read source
- [3] Baseten model library — Aleph 2.0 (dedicated deployment listing, offered for deployment inquiries)Read source
- [4] Runway Dev — "API Pricing & Costs" (credits at $0.01 each; Gen-4.5 at 12 credits per second and Aleph 2.0 at 28 credits per second, i.e. $0.12 and $0.28 per second)Read source
- [5] Unite.AI — "Baseten Adds Runway's Gen-4.5 and Aleph 2.0 to Dedicated Inference", 14 October 2026 (single-tenant documentation details, including isolated VPC and private networking options, and Enterprise custom pricing)Read source