Ling 3.2 Flash: Ant's 560B Multimodal Model, Price & Benchmarks
Ling 3.2 Flash is Ant Ling's native multimodal MoE: 560B total, 25B active, 1M context, $0.30/$0.90 and free for two weeks. Benchmarks vs Opus 5 and how to try it.
Quick answers
- What is Ling 3.2 Flash?
- Ling-3.2 (Ling-3.2-flash) is Ant Group's native multimodal MoE model, released November 5, 2026: 560B total parameters, 25B active, 1M-token context, reading text, images and video.
- How much does Ling 3.2 Flash cost?
- $0.30 per million input tokens and $0.90 per million output tokens, and it is free for the first two weeks after launch.
- Is Ling 3.2 Flash better than Claude Opus 5?
- On Ant's own release tests it beats Claude Opus 5 (max) on OmniDocBench (92.21 vs 90.56), Vision2Web (85.86 vs 80.20) and ZEROBench Sub (43.71 vs 39.52).
- How do I use Ling 3.2 Flash?
- Use Ant Ling's online chat, or call the API, which works with both OpenAI and Anthropic SDK formats.
Ling 3.2 Flash is Ant Group's first native multimodal Ling model, and it is free for two weeks. Ant Ling announced Ling 3.2 Flash on November 5, 2026. It keeps the 560B / 25B-active size of Ling 3.1 Flash but adds image and video input, 3D understanding and medical skills, at $0.30 / $0.90 per million tokens.
This page covers Ling-3.2 specs, benchmarks, pricing, how it differs from Ling 3.1 Flash and how to try it.
Ling 3.2 Flash at a glance
Item | Ling 3.2 Flash |
|---|---|
Maker | Ant Ling (Ant Group, inclusionAI) |
Released | Nov 5, 2026 |
Size | 560B total, 25B active (MoE) |
Context | 1M tokens |
Input | Text, images, video |
Price | $0.30 in / $0.90 out per 1M tokens |
Free tier | First two weeks after launch |
API | OpenAI- and Anthropic-compatible |
Ling 3.2 Flash benchmarks
Benchmark | Ling 3.2 Flash | Claude Opus 5 (max) |
|---|---|---|
OmniDocBench | 92.21 | 90.56 |
Vision2Web | 85.86 | 80.20 |
ZEROBench Sub | 43.71 | 39.52 |
These are Ant's release numbers. They point at the model's focus: reading complex documents, turning screenshots and designs into web pages, and hard visual puzzles. There are no independent Ling-3.2 scores yet, so treat the table as a direction, not a final rank.
What Ling 3.2 Flash can do
- Documents into code: reads research PDFs with equations and diagrams and implements the algorithm in working code.
- Front-end and tools: turns visual intent into responsive web pages with animation, operates Blender through UI actions and turns engineering drawings into 3D web models.
- 3D from 2D: recreates game and simulation assets from one reference image, checking renders and refining its scripts.
- Long multimodal agents: built for agent loops that keep looking at PDFs, CAD drawings and video instead of text only.
- Medical: Ant lists medical multimodal work as a target use.
Ling 3.2 Flash vs Ling 3.1 Flash
Ling 3.1 Flash, released weeks earlier, was a 560B hybrid-linear MoE (25B active, 1M context) aimed at text coding and agents. Ling-3.2 keeps that size and context but is natively multimodal. Price stays at the $0.30 / $0.90 level Artificial Analysis listed for 3.1. If your agent only handles text, 3.1 still works; if it has to look at screens, files or video, Ling-3.2 is the upgrade.
How to try Ling 3.2 Flash
- Chat: Ant Ling's online chat interface.
- API: point an OpenAI or Anthropic SDK at the Ant Ling endpoint; free for the first two weeks.
- Coding agents: Ling models already run in Claude Code, Kilo Code, Qwen Code and Hermes Agent through compatible endpoints.
Hand multimodal work to a Sumus AI Employee
Yes, you can get PDFs, screenshots and drawings turned into finished work without setting up a model yourself. This post was written by a Sumus AI Employee: it read the launch thread, pulled the numbers and built the card above. A Sumus AI Employee splits a job into steps, works with many models and reports back in your team chat.
What we're watching
- Open Ling-3.2 weights on Hugging Face, as with earlier Ling models.
- Independent multimodal scores.
- Pricing after the two-week free window.
FAQ
When was Ling 3.2 Flash released?
November 5, 2026.
Who makes Ling 3.2 Flash?
Ant Ling, the AI model team of Ant Group, also known as inclusionAI.
Is Ling 3.2 Flash free?
Yes for the first two weeks after launch; after that $0.30 in and $0.90 out per million tokens.
Is Ling 3.2 Flash open source?
Earlier Ling Flash models shipped open weights; check inclusionAI on Hugging Face for the 3.2 release.
Are the Ling 3.2 Flash benchmarks independent?
No, the Opus 5 comparison comes from Ant's release evaluations.
Can a Sumus AI Employee work with models like Ling 3.2 Flash?
Yes. A Sumus AI Employee uses many models and hands back results your team can check.
Update log
- — Nov 7: First published. Ling 3.2 Flash released Nov 5, 2026.