Advertise
AI

Ant Group releases Ling-3.2-flash, claims edge over Claude Opus 5

The 560-billion-parameter model handles text, images and video with up to 1 million tokens of context, and its API is free for two weeks.

Ant Group released the multimodal AI model Ling-3.2-flash on November 5, 2026, saying it outperforms Anthropic's Claude Opus 5 on several vision and document benchmarks, according to aitimes.com.

The model has 560 billion total parameters and activates 25 billion in inference through a mixture-of-experts structure. It handles up to 1 million tokens of context and processes text, images and video together.

In Ant Group's own measurements, Ling-3.2-flash scored 92.21 on OmniDocBench 1.5, against 90.56 for Claude Opus 5. It also posted 43.71 on certain ZEROBench visual reasoning items, versus 39.52 for Claude Opus 5, and 85.86 on Vision2Web, against 80.20.

External ranking and pricing

DesignArena, an external evaluation platform, placed Ling-3.2-flash 10th on its Image to Frontend Arena with an Elo of 1238. It lists pricing at $0.30 per 1M input tokens and $0.90 per 1M output tokens.

Ant Group is offering the API free for two weeks after launch and has not disclosed official pricing afterward. The model supports the API call formats of OpenAI and Anthropic, and is available through an API and an online chat service. The company has not announced plans to release the model's weights.

Ant Group, which operates Alipay, has stepped up in-house AI model and agent development in recent years.

Quick answers

How much does Ling-3.2-flash cost?

DesignArena lists $0.30 per 1M input tokens and $0.90 per 1M output tokens. Ant Group is offering the API free for two weeks after launch and has not disclosed official pricing afterward.

How large is Ling-3.2-flash?

It has 560 billion total parameters and activates 25 billion in inference through a mixture-of-experts structure, with support for up to 1 million tokens of context.

How do I access Ling-3.2-flash?

It is available through an API that supports the call formats of OpenAI and Anthropic, and through an online chat service.

Source