Artificial Analysis Launches Code2Video-AA Benchmark for Marketing Video Generation

1 reports 1 reporting sources 2 hours agoNew

Understand the event

AI overview

Artificial Analysis introduced Code2Video-AA to evaluate models that generate marketing videos in code, with Claude Opus 5.5 leading its rankings and fourth-place GPT-6.1 Sol costing about one-tenth as much per task. Artificial Analysis reported that Claude Opus 5.5 (max with fallback) scored 1536 Elo at $3.08 per task, while GPT-6.1 Sol (max) scored 1369 at $0.31 per task—a 167-point gap. Built in partnership with HeyGen, Code2Video-AA is Artificial Analysis’s independent implementation of HeyGen’s Code2Video Bench. It evaluates 27 models on 114 product and brand marketing briefs using HyperFrames HTML compositions, with more than 35,000 verdicts across five judging axes. Each model gets a single response without tools, and rendering runs without network access, so the rankings apply to generation under those constraints.

Generated by AI from reports · Updated 1 hours ago

Latest development10/27 06:22 (UTC+8)

Artificial Analysis introduces Code2Video-AA to benchmark coded marketing video generation Artificial Analysis · X

Models in this story

Claude Opus 5.5#1 overall$4 input / $20 output per million tokens
GPT-6.1 Sol#7 overall$2 input / $10 output per million tokens

Compare their API prices API prices of all models

Report timeline

Follow the reports to explore different sides of the event.

10/27

06:22
Artificial Analysis · X
Artificial Analysis introduces Code2Video-AA to benchmark coded marketing video generation
Artificial Analysis introduces Code2Video-AA, its independent implementation of HeyGen’s Code2Video Bench, built in partnership with HeyGen to evaluate models generating marketing videos in code. It tests 114 briefs using HyperFrames HTML compositions, with over 35,000 verdicts ranking 27 models across five judging axes. Claude Opus 5.5 leads with 1536 Elo at $3.08 per task, while fourth-place GPT-6.1 Sol scores 1369 at $0.31 per task; models receive a single response without tools, and rendering has no network access.

Model releases, price changes and compute developments, with original sources. Free, confirmed by email, and easy to unsubscribe. Read a past issue

Event heat trend

There is not enough continuous observation data to plot a trend yet.