Runway Gen-4.5: What It Is, What It Costs, Its Limits
Runway released Gen-4.5 on December 1, 2025, and launched it at the top of a public text-to-video leaderboard. The model turns a text prompt, or an image plus a prompt, into a short video clip. Runway's API docs still call it the company's state-of-the-art text and image to video model. You can use it in the app or through the API. Below: what Gen-4.5 does well, its price per second, and the limits Runway admits. Each new clip sees only your prompt, so the last section covers keeping a consistent look across generations.
What is Runway Gen-4.5, in one sentence?
Runway Gen-4.5 is a video generation model that turns a text prompt, optionally with a first-frame image, into a short clip with realistic motion.
Runway's announcement calls it the world's top-rated video model. That claim rested on the Artificial Analysis text-to-video leaderboard as of November 30, 2025, where Gen-4.5 scored 1,247 Elo. Veo 3 had 1,226 and Kling 2.5 had 1,225. Sora 2 Pro had 1,206. The rankings come from blind human preference votes across thousands of pairwise comparisons. Runway says it built and serves the model entirely on NVIDIA GPUs, with inference on Hopper and Blackwell chips.
How does Gen-4.5 compare with other Runway models?
The ranking has moved since launch. Artificial Analysis now runs a newer version of its text-to-video leaderboard. When we checked it in October 2026, Wan 3.0 led with 1,174, and Gen-4.5 was not in the top 10. Scores from the two leaderboard versions are not directly comparable.
| Gen-4.5 | Gen-4 Turbo | Aleph 2.0 | |
|---|---|---|---|
| Made by | Runway | Runway | Runway |
| Input | Text, or image plus text | Image to video | Video to video (editing) |
| API price | $0.12 per second | $0.05 per second | $0.28 per second |
| App price | 12 credits per second | Not compared here | Not compared here |
| Use it for | New shots from a prompt | Cheap animation of a still | Restyling or editing footage |
Sources: API prices from Runway's API model list, app credits from Runway pricing, Aleph 2.0 from Runway's changelog.
The API model list also carries third-party models such as Veo 3.1 and Wan 3.0. You can call a newer leader through the same Runway API key.
What does Gen-4.5 do well?
Runway's announcement highlights physical accuracy. Objects are supposed to move with believable weight and momentum, and liquids are supposed to behave like liquids. It also claims strong prompt adherence and range from photorealistic to stylized footage. These are Runway's own claims. The Elo figure is the only third-party measure, and it comes from preference votes. We found no published physics test.
Gen-4.5 launched as text to video. On January 21, 2026, Runway's changelog added image to video, so you can supply a first-frame image with the prompt. Since August 10, 2026, Gen-4.5 output can be exported as ProRes 4444 or PNG sequences.
What are Gen-4.5's known limitations?
Runway lists three failure modes in its own announcement:
- Causal reasoning. Effects sometimes come before their causes, so a door can swing open before the hand touches it.
- Object permanence. Objects can vanish or appear between frames.
- Success bias. Actions succeed more often than they should. A thrown ball tends to go in.
The door example in point 1 is ours. Plan for these when you storyboard. Keep shots short and check continuity across cuts. Generate several takes of any shot where a cause and effect must read clearly.
What does Gen-4.5 cost?
In the Runway app, Gen-4.5 costs 12 credits per second of video, so a 5-second clip is 60 credits. The pricing page lists these plans:
- Free: $0, with 125 one-time credits
- Standard: $12 a month billed annually ($15 monthly), 625 credits a month
- Pro: $28 a month billed annually ($35 monthly), 2,250 credits a month
- Max: $76 a month billed annually ($95 monthly), 9,500 credits a month
At 12 credits a second, Standard's monthly credits cover about 52 seconds of Gen-4.5 video before retakes. The API charges $0.12 per second. A 5-second clip there costs $0.60.
How do you keep a consistent style across generations?
Each Gen-4.5 call is independent. The model sees the prompt and the optional first frame you send, and nothing from your earlier clips.
Inside the Runway app, the newer Runway Agent adds persistent context. Its product page says Brand Kits keep your logos and style references available in every session. Workflows repeat production steps on later projects, and Skills save your team's know-how. Per the changelog, Agent sessions have lived inside Projects with access to their assets since August 11, 2026.
Through the API, you carry that context yourself. Reuse the same first-frame image with a fixed prompt template for camera and lighting. Note which wordings and frames produced the shots you kept.
The bottom line
Use Gen-4.5 for short shots from a prompt or a still. Budget 12 credits or 12 cents per second before retakes. Newer, higher-ranked models now sit in the same API, so check the current leaderboard before you commit. Write your style rules down once, as a prompt template plus reference frames, and reuse them on every call.
How do you keep creative decisions when an agent drives the API?
You can also script video generation from a coding agent. Claude Code or Codex writes the API calls and retries failed shots. Over a project, that agent finds out which prompt phrasing kept the palette stable and which reference frame fixed a character's face. Gen-4.5 keeps none of it, and Runway's Brand Kits live inside the Runway app, out of reach of an agent working through the API.
memU keeps that history as readable Markdown the agent can retrieve. A scheduled background task slices new session logs into jobs. The agent distills each job into memory and skill Markdown, such as a "house video style" skill. memU commits the result. memU makes no LLM or chat calls itself. It stores, embeds and retrieves the Markdown the agent prepared. A standing instruction in CLAUDE.md or AGENTS.md tells the agent to retrieve before answering. All hosts share one backend, so a style rule learned in Codex is available in Claude Code. memU does not connect to Runway or change how Gen-4.5 generates.
The hosted version is free and syncs across devices. Get an API key from memu.so and send your agent one message:
Read https://memu.pro/SKILL.md, follow its instructions to install and configure memU, API Key is
YOUR_MEMU_API_KEY.
Self-hosting and the list of supported agents are in the memU GitHub repository.
FAQ
Is Runway Gen-4.5 free?
Runway's Free plan gives 125 one-time credits, but the pricing page does not say whether Gen-4.5 is among its models. Runway's announcement says Gen-4.5 is available on all paid plans.
How much does Gen-4.5 cost per second?
Gen-4.5 costs 12 credits per second in the Runway app and $0.12 per second through the Runway API. A 5-second clip is 60 credits or $0.60.
Can Gen-4.5 do image to video?
Yes. Runway added image to video on January 21, 2026. You supply a first-frame image along with your text prompt.
Is Gen-4.5 still the best AI video model?
It led the Artificial Analysis text-to-video leaderboard at launch with 1,247 Elo. On the newer version of that leaderboard in October 2026, Wan 3.0 led. Gen-4.5 was not in the top 10.
Does Gen-4.5 remember earlier generations?
No. Each generation uses only the prompt and optional image you send. For consistency, reuse reference frames with a fixed prompt template. Inside the Runway app, Brand Kits in Runway Agent do this for you.