AnswersPublished24 sources
What did Ant Group announce about Ling 3.2 flash on November 6, 2026, including its native architectural integration of image, text and video capabilities, 560 billion total and 25 billion active parameters, 1 million token context window, mixture of experts efficiency and compar
Ant Group announced Ling 3.2 flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1 million token context window, emphasizing visual reasoning and multimodal agents. [6] November 6 reporting describes the announcement as occurring that day, although other launch...
Published byImages generated with GPT Image 2
Research answer
Key takeaways
- Ant Group announced Ling 3.2 flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1 million token context window, emphasizing visual reasoning and multimodal agents.
- [6] November 6 reporting describes the announcement as occurring that day, although other launch coverage dates it November 5; the available evidence does not resolve that discrepancy.
- [8][10] Architecture and intended uses Native multimodality: Ant describes image, text and video capabilities as integrated at the architectural level, and its OpenAI compatible API explicitly accepts all three input types.
- [14][4] Efficiency: The reported mixture of experts design activates approximately 25 billion parameters per token rather than the entire 560 billion parameter model, preserving model capacity while limiting computation per step.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did Ant Group announce about Ling 3.2 flash on November 6, 2026, including its native architectural integration of image, text and vide. Article summary: Ant Group announced Ling 3.2 flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1 million token context window, emphasizing visual reasoning and multimodal agents.. Topic tags: general web, openai, ai, workflow, productivity. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, cha
Ant Group announced Ling-3.2-flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1-million-token context window, emphasizing visual reasoning and multimodal agents. 6 November 6 reporting describes the announcement as occurring that day, although other launch coverage dates it November 5; the available evidence does not resolve that discrepancy.
8
10
Architecture and intended uses
- Native multimodality: Ant describes image, text and video capabilities as integrated at the architectural level, and its OpenAI-compatible API explicitly accepts all three input types.
14
4
- Efficiency: The reported mixture-of-experts design activates approximately 25 billion parameters per token rather than the entire 560-billion-parameter model, preserving model capacity while limiting computation per step.
- Focus areas: Complex visual understanding and reasoning, long-horizon multimodal agents, 3D understanding and interaction, and medical multimodal tasks are the four stated priorities.
- Earlier models: Secondary coverage characterizes Ling-3.2-flash as retaining Ling-3.1-flash’s 560-billion-parameter scale while adding native multimodal capabilities; Ling-3.0-flash-VL already supported image and video input.
11
Performance claims
Ant-reported results put Ling-3.2-flash ahead of Claude Opus 5 (max) on these three benchmarks; these are vendor-reported comparisons, not independently verified test results. 13
8
| Benchmark | Ling-3.2-flash | Claude Opus 5 (max) |
|---|---|---|
| OmniDocBench 1.5 | 92.21 | 90.56 |
| ZEROBench subset | 43.71 | 39.52 |
| Vision2Web | 85.86 | 80.20 |
The scores above are reported in launch coverage. 13
14
- DesignArena: Launch reporting places it in the top 10 of the image-to-frontend arena with an Elo score of 1,238.
8
- Relative weaknesses: Insufficient evidence to verify the requested comparisons with GPT-5.6-Sol (max), Claude Opus 5 (max) and Kimi K3 (max) on Humanity’s Last Exam’s multimodal subset, BabyVision and MathVision.
Pricing, access and open weights
- Launch offer: Ant’s official pricing documentation confirms free access for the first two weeks after launch, followed by a return to regular pricing.
- Dollar pricing: One supplied report gives $0.30 per million input tokens and $0.90 per million output tokens, but other coverage says post-promotion rates were unpublished; the official excerpt supplied here does not establish those dollar rates.
15
- API access: Availability through Ant’s Ling API and OpenAI compatibility are confirmed by official documentation; official Claude Code instructions and secondary coverage also support Anthropic-compatible access.
- Open-source weights: Secondary coverage describes Ling-3.1-flash as having MIT-licensed weights, while presenting Ling-3.2-flash’s weights as unreleased at launch.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "What did Ant Group announce about Ling 3.2 flash on November 6, 2026, including its native architectural integration of image, text and video capabilities, 560 billion total and 25 billion active parameters, 1 million token context window, mixture of experts efficiency and compar"?
Ant Group announced Ling 3.2 flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1 million token context window, emphasizing visual reasoning and multimodal agents.
What are the key points to validate first?
Ant Group announced Ling 3.2 flash as a native multimodal flagship with 560 billion total parameters, 25 billion active parameters and a 1 million token context window, emphasizing visual reasoning and multimodal agents. [6] November 6 reporting describes the announcement as occurring that day, although other launch coverage dates it November 5; the available evidence does not resolve that discrepancy.
What should I do next in practice?
[8][10] Architecture and intended uses Native multimodality: Ant describes image, text and video capabilities as integrated at the architectural level, and its OpenAI compatible API explicitly accepts all three input types.
Sources
aicoder.comAnt Ling Launches Ling-3.2-flash: 560B/25B Native Multimodal ...