Runway Gen-4.5

Artificial Analysis identifies Gen-4.5 as its 2025 year-end video-model leader, about 200 Elo ahead of prior leader Sora, supporting a high technical-capability score for its intended medium. Official Runway materials document API model ID `gen4.5` and both text-to-video and image-to-video generation. This is strong modality-specific evidence, not evidence of broad language intelligence, coding, or agent operation. Runway publishes an operational unit price of 12 credits per generated second, with clearly stated surcharges for professional formats and HDR/10-bit. That makes pricing more transparent than Pegasus 1.5’s low-clariy anchor, although the supplied evidence does not convert credits to currency or establish output-quality-per-dollar. API documentation supports a solid developer-experience score; no supplied latency, uptime, throughput, or regional-availability measurements justify a higher availability score. Gen-4.5 exceeds generalist anchors such as GPT-5.6 Terra on video-specific technical evidence, but remains materially below Claude Opus 4.8, GPT-5.5, and Kimi K2.5 for coding and reasoning because no applicable results exist. It is similar to Gemini 3.5 Flash in input modality breadth but narrower: its documented output is video. Public preference is mixed: Text-to-Video Arena records 43,610 votes but ranks it 27th (1224±9), tempering the year-end Artificial Analysis result. Evidence quality is good for video capability and pricing, but sparse outside that scope.