Tasks and inputs
Ling 3.2 Flash is listed for Image-text-to-text, Multimodal, Text → Text, Video-text-to-text. It accepts text, image, video and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Ling 3.2 Flash is inclusionAI's 560B-A25B multimodal MoE with a 1M-token window, built for visual agents, 3D tasks and medical image reasoning.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Free for the first two weeks after the 5 November 2026 launch; no paid per-token rate published at the time of writing. Indicative only — confirm on your own Ant Ling account.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
Teams that already route work to Ling 3.1 Flash will notice the jump straight away: this release reads images and video natively instead of treating vision as a separate model. It is a sparse mixture-of-experts with 560B total parameters and 25B active per token, and it accepts text, image and video input with text output.
The intended workloads are unusually specific. Complex visual understanding, long-horizon multimodal agents, 3D understanding and interaction, and medical multimodal tasks are the four areas inclusionAI calls out. That list tells you where to point your first tests.
| Field | Value |
|---|---|
| API model ID | Ling-3.2-flash |
| Architecture | Native multimodal MoE, 560B total / 25B active |
| Context window | 1,000,000 tokens |
| Input | Text, image, video |
| Output | Text |
| Reasoning | Hybrid; can be switched off per request |
| Tool calling | Supported |
| Open weights | None published at launch |
| Launch | 5 November 2026 on Ant Ling's developer platform |
Access is through Ant Ling's developer platform, which exposes the model behind both OpenAI-compatible and Anthropic-compatible APIs, plus a chat interface. In practice that means most existing client code needs only a base URL and model-name change. Launch terms made the model free for the first two weeks; no paid rate had been published when we checked, so treat any figure you see elsewhere as provisional and confirm it in your own console.
Listings disagree on two details. The launch date appears as 5 November on Ant Ling's changelog, while one aggregator snapshot is dated 13 November, which most likely marks when that route went live rather than a different model. Maximum output is the bigger issue: one catalogue gives 32,768 tokens, another 262,144. Send a long-generation request with an explicit max_tokens and read the cap from the error or the usage block before you design around either number.
The family now spans several shapes. Ling 3.0 Flash VL was the earlier vision-capable option with downloadable weights, and Ling 3.0 Flash remains text-only. Choose 3.2 when one model must handle screenshots, clips and long instructions together; stay on an open checkpoint if self-hosting is a hard requirement.
A sensible trial:
Browse the wider inclusionAI catalogue or the multimodal models list to line it up against alternatives.
Compare the model on the work you need to complete.
Ling 3.2 Flash is listed for Image-text-to-text, Multimodal, Text → Text, Video-text-to-text. It accepts text, image, video and produces text. Use the publisher’s input format and preprocessing for your chosen deployment.
Documented limits: 1M context. Measure a completed task using the prompts, documents and tools your application will use. Include reasoning and retries in the result.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
NVIDIA
Alpamayo 2 Edge is NVIDIA's open 2B-class driving VLA that outputs trajectories with causal explanations, sized for DRIVE AGX Thor in-vehicle use.
shanghai-ai-lab
Sage-SciPerception is Shanghai AI Lab's set of projectors feeding crystals, spectra, biological sequences and biosignals into a language model's input.
Swiss AI
Apertus 1.5 8B is the Swiss AI Initiative's open multilingual model, now taking text, images and audio with a 262K-token context window.
kyutai
REKI-2B is Kyutai's open 2.4B robot vision-language-action model with cheap register-token memory, pretrained only on public data.
black-forest-labs
FLUX 3 Action is Black Forest Labs' open 7B world action model for robots, predicting action chunks and future video from cameras, state and text.