Runway Gen-4.5: The First General World Model
Stop generating video. Start simulating reality. Outperforming Google Veo and OpenAI Sora 2 Pro in blind tests, Gen-4.5 introduces causal understanding, true physics, and unbroken world consistency. Code name 'David'—because precision beats size.
Click to try these prompts
#1 on Video Arena Leaderboard
The industry has spoken. In independent blind tests conducted by Artificial Analysis, Gen-4.5 debuted with a record-breaking Elo score of 1247, statistically outperforming Google Veo 3 and OpenAI Sora 2 Pro. It isn't just an upgrade; it is the new standard for visual fidelity.
Simulation, Not Just Generation
Previous AI models hallucinated motion. Gen-4.5 simulates it. With a deep understanding of causal logic, objects in Gen-4.5 obey the laws of physics. Gravity feels heavy, collisions have impact, and fluids flow naturally. If a glass drops, it doesn't morph—it shatters because the model understands *why*.
Unbroken Narrative Continuity
Filmmaking requires more than random clips. It requires consistency. Gen-4.5 excels at 'World Consistency,' maintaining character identity, environmental geometry, and lighting styles across multiple generated shots. Tell a cohesive story with the same actor in the same world, from different angles.
Mastering Gen-4.5: From Prompt to Premiere
Leverage the power of the 'David' model in 4 steps
1. Describe the Physics
Don't just describe the visual; describe the action. Gen-4.5 understands verbs like 'collide', 'flow', 'shatter', and 'bounce'. Be specific about weight and momentum to trigger the physics engine.
2. Set World Consistency
Upload a reference image or use a character seed. Gen-4.5 locks onto the subject's identity, allowing you to generate multiple angles of the same actor without the 'morphing' issues of older models.
3. Direct the Camera
Use professional camera terminology. Prompts like 'Rack focus', 'Dolly zoom', and 'Low-angle tracking shot' are natively understood, giving you directorial control over the virtual lens.
4. Generate in Native HD
Gen-4.5 is optimized for high-definition output on NVIDIA Blackwell architecture. Generate crisp, broadcast-ready video directly, reducing the need for external upscaling tools.
Creators vs. Goliaths 🛡️
See why independent creators and studios are switching to Runway Gen-4.5.
"The physics in Gen-4.5 are terrifyingly good. I asked for a car crash, and the metal crumpled exactly how it should. It understands material properties in a way Sora just doesn't."
"We finally have 'World Consistency'. Being able to keep my main character looking like herself across 5 different shots changes everything for our storyboard animatics."
"The 'David' codename is fitting. Runway's team is a fraction of Google's size, yet Gen-4.5 just beat Veo 3 on the Video Arena leaderboard. The efficiency of this model is incredible."
"No more 'dream-like' floating objects. When I put a cup on a table in Gen-4.5, it sits there. The groundedness of the motion makes it usable for actual narrative work."
"I use it for texture reference and fluid sims. The water and smoke effects are close to what I'd get from a Houdini render, but generated in seconds."
"Speed matters. Running on NVIDIA Blackwell means we iterate faster. The HD output is crisp enough that we're using it directly in pitch decks without retouching."
Pricing Plans
Choose a plan that fits your needs. All plans include basic features and support.
🔥 Limited Time Trial Offer!
Start creating with AI for just $6.99
100 credits
Secure payment by Stripe
Basic Monthly
Perfect for individuals and light users
500 credits/month
Includes
- 500 credits/month
- Up to 50 videos/month
- Up to 500 images/month
- 3 Parallel tasks
- All-in-one multi-model support
- Text/Image/Video to video
- Text/Image/Chat to image
- 100+ AI image & video tools
- Faster generation speed
- No-watermark outputs
- Copy protection
- Priority support
Billed monthly. Cancel anytime.
Pro Monthly
For professional creators and teams
1,500 credits/month
Includes
- 1500 credits/month
- Up to 150 videos/month
- Up to 1500 images/month
- 3 Parallel tasks
- All-in-one multi-model support
- Text/Image/Video to video
- Text/Image/Chat to image
- 100+ AI image & video tools
- Faster generation speed
- No-watermark outputs
- Copy protection
- Priority support
For professional creators
Max Monthly
Designed for large enterprises and professional studios
3,500 credits/month
Includes
- 3500 credits/month
- Up to 350 videos/month
- Up to 3500 images/month
- 3 Parallel tasks
- All-in-one multi-model support
- Text/Image/Video to video
- Text/Image/Chat to image
- 100+ AI image & video tools
- Faster generation speed
- No-watermark outputs
- Copy protection
- Priority support
For enterprise-level users
Frequently Asked Questions about Gen-4.5
Everything you need to know about the new 'David' Model
Gen-4.5 (codenamed 'David') represents a shift from 'Video Generation' to 'General World Models'. The key difference is its understanding of physics and causality. Unlike competitors that often create dream-like, morphing visuals, Gen-4.5 respects gravity, collision, and object permanence, resulting in significantly more realistic and grounded motion. It is currently ranked #1 on the Video Arena leaderboard.
Start Simulating the World
Join the new era of AI video. Try the model ranked #1 for realism and physics.