CL-SN55 941 —
GPT-61S 885 -0.6%
GPT-56T 861 —
GPT-6S 825 -1.4%
CL-OP55X 819 -1.4%
MUSE-SPK 793 -1.4%
QWEN-38X 779 -2%
GPT-6A 778 -0.5%
CL-OP5H 712 -1.7%
CL-OP5X 710 -1.7%
KIMI-K3X 700 -2.5%
CL-OP46H 695 -1.7%
CL-FAB5H 693 -1.7%
CL-OP55H 693 -1.7%
CL-OP47H 684 -1.7%
GEM-37FH 649 -1.8%
GEM-38FH 634 -3.4%
GPT-56S 574 -0.5%
CL-OP47 540 -0.6%
INKL 531 —
GEM-3P 494 —
CL-OP46 490 —
CL-OP48 483 —
GEM-31P 459 -5.4%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 885 -0.6%
GPT-56T 861 —
GPT-6S 825 -1.4%
CL-OP55X 819 -1.4%
MUSE-SPK 793 -1.4%
QWEN-38X 779 -2%
GPT-6A 778 -0.5%
CL-OP5H 712 -1.7%
CL-OP5X 710 -1.7%
KIMI-K3X 700 -2.5%
CL-OP46H 695 -1.7%
CL-FAB5H 693 -1.7%
CL-OP55H 693 -1.7%
CL-OP47H 684 -1.7%
GEM-37FH 649 -1.8%
GEM-38FH 634 -3.4%
GPT-56S 574 -0.5%
CL-OP47 540 -0.6%
INKL 531 —
GEM-3P 494 —
CL-OP46 490 —
CL-OP48 483 —
GEM-31P 459 -5.4%
GPT-6L 436 —
← Back to feed

Claude Opus 5.5 Completes a Full Explainer Video in 30 Minutes in One-Shot Demonstration

launchvideo.io, a web application that generates explainer videos from a URL or prompt by running Claude Opus 5.5 on serverless browser rendering agents, reached Hacker News’ front page on September 25. The post drew attention to a capability that had been circulating in independent tests: Opus 5.5 handling multi-step creative production tasks at a quality and speed that earlier models could not.

A separate YouTube demonstration — titled “INSANE ONE-SHOT! Opus 5.5 made this entire video in 30 min” — showed the model producing a complete video in a single session without revision passes. The creator noted the rough edges were intentionally left in to show what unassisted generation looked like. There were few.

The Codebase Migration Numbers

The video production capability is a visible proxy for something quantified more precisely in Anthropic’s announcement. An early access tester used Opus 5.5 to audit and fix a 200,000-line codebase in under three hours. The same job took Opus 5 more than 20 hours and consumed 2.5 times as many tokens.

Anthropic’s own internal test ran a different benchmark: a C-to-Rust port of the HAProxy load balancer. The early tester’s codebase result is one data point, but the directional signal matches what the architecture suggests — Opus 5.5 is optimised for long, sprawling jobs with high context demands.

3D Modelling Claims

Independent reviewer Peter Yang, testing Opus 5.5 on visual design tasks including 3D generation and anime creation, rated its output on par with GPT-6 Astra. His characterisation: “they have solved 3D modelling.”

That framing outruns the underlying capability in its literal reading. What the comparison shows is that two frontier models are now producing visual outputs at a threshold for professional usability in specific creative workflows. Whether that constitutes a “solved” problem depends on what benchmark you use to define solved.

What the Pattern Means

The launchvideo.io submission is notable for what it is not: marketing produced by Anthropic. It is a user-built product submitted directly to HN, where the community can scrutinise the output. The reach suggests the quality was there.

Opus 5.5 launched three days ago at $4 per million input tokens and $20 per million output. Community tests of the explainer video workflow reported inference costs around $3 to $4 per generation. At that price, production-quality video from a prompt is within reach of any API subscriber.

The codebase migration benchmark remains the harder signal. Long-horizon agentic coherence — holding a complex system in context, making thousands of related decisions correctly, and delivering a finished output — is the capability that separates models ready for unsupervised production work from those that still require a human in the loop.

Opus 5.5’s early results suggest it has crossed that threshold in more than one domain.