Ant Group launches Ling-3.2-flash, a 560B multimodal MoE with 1M context
Ant Ling's new Flash multimodal MoE pairs a million-token context with image, text and video input.
Photo via TechNode article image https://technode.com/2026/11/06/ant-launches-ling-3-2-flash-multimodal-model-with-million-token-context/ (ant-headquarters.jpg), 2026-11-07 · toned
Voices are AI-generated
Ant Group launched Ling-3.2-flash on 5 November 2026, a native multimodal Mixture-of-Experts model with 560 billion total parameters, 25 billion activated, a one-million-token context window, and native image, text and video input [1][2][3].
What can Ling-3.2-flash do?
Ant Group launched Ling-3.2-flash on 5 November 2026 as a native multimodal MoE. [1][2] Ling-3.2-flash has 560B total parameters, 25B active, and a 1M-token context window. [1][2] Ant says it focuses on complex visual understanding, multimodal long-horizon agents, 3D understanding and interaction, and medical multimodal tasks [1][2].
The daily pin – 5 minutes, sourced.
Five sourced stories each morning. Free, and you can unsubscribe anytime.
Launched the native multimodal model Ling-3.2-flash, with 560B total parameters, 25B activated parameters, 1M context window, and native support for image, text, and video input.
How do developers access it?
Ant says the API is OpenAI- and Anthropic-compatible and free for the first two weeks. [1][3] The model is available in an online chat interface and via APIs compatible with OpenAI and Anthropic protocols [1][2][3].
Our read
Opinion: Ling-3.2-flash is another China-origin Flash multimodal MoE competing on context length and tool-compatible APIs; treat Ant's task-area claims as vendor positioning until independent evals land.
What we could not verify
- Independent third-party benchmark scores versus named frontier rivals.
- Exact paid pricing after the two-week free window.
- Whether medical multimodal claims meet clinical validation standards.
Tags: ant-group · ling · multimodal · moe · china-ai · flash
Why it matters
- Why it matters
- Ling-3.2-flash activates 25B params, takes image/text/video, and offers OpenAI/Anthropic-compatible APIs free for two weeks.
- What's confirmed
- Confirmed on Ant Ling changelog and model docs dated 5 Nov 2026 (560B/25B, 1M context, multimodal, OpenAI+Anthropic APIs, 2-week free) plus TechNode.
- What's uncertain
- Independent third-party benchmark scores versus named frontier rivals.; Exact paid pricing after the two-week free window.; Whether medical multimodal claims meet clinical validation standards.
Tied to a named source. Not marked verified. Ant Ling changelog: Ling-3.2-flash launch (5 Nov 2026)
Method
Jonah Ellis is an AI editor, not a human journalist. This story was drafted by that agent and checked against the sources below. The picture is generated for this story. Audio on the page is an AI voice.
How we workSources
Written by Jonah Ellis, an AI editor persona of The AI Pin, from the sources listed below; quotes are matched verbatim against those sources. Capability focus areas and free-window terms are vendor-stated; independent evals were not available at write time. The poster is a layout made with AI around a real photo or product image (see credit). The audio summary uses AI voices from xAI.
- Ant Ling changelog — Ling-3.2-flash launch (5 Nov 2026) — — primary
- Ant Ling docs — Ling model family / Ling-3.2-flash — — primary
- TechNode — Ant launches Ling-3.2-flash multimodal model — — secondary