Ling 3.2 Flash Benchmarks, Specifications & Availability

Ling 3.2 Flash

Explore Ling 3.2 Flash from inclusionAI: published specifications, source-linked vendor benchmarks, independent evaluator coverage and recorded pricing when available.

Published specifications

Provider: inclusionAI
Access: Proprietary
License: Proprietary
Context window: 256K
Total parameters: 560B
Active parameters: 25B
Released: 2026-11-05
Modalities: text, image, video
Family: Ling 3.2

Announcement · Website

Model notes

Ant Ling’s native multimodal model, announced by AntLingAGI on 5 November 2026 (19:40 UTC) with 560B total / 25B active parameters and a 14-panel results chart. Hosted on Computrix, whose page lists API ID ling-3.2-flash, release date 5 November 2026, 256K context, text, image and video input with text output, and $0 / $0 per million tokens from a single provider (Ant Digital Technologies); that is a marketplace listing, not an OpenRouter tariff, so no price row. No official weights at the 6 November 2026 check: inclusionAI/Ling-3.2-flash does not resolve on Hugging Face and inclusionAI lists no Ling 3.2 repository, so access is closed and the license is unpublished. No OpenRouter route. Not linked to Ling 3.0 Flash VL or Ling 3.1 Flash as an upgrade: Ant names it a new model, not a checkpoint of either. Launch scores have no stated reasoning effort; the peers’ legend says max.

Compare Ling 3.2 Flash with other models →Compare subscription plans Explore data coverage

Pricing · OpenRouter

No OpenRouter pricing is recorded for this model. Missing rates are not free.

Official / vendor benchmarks

Default headline records. Own-vendor, peer-vendor and third-party provenance remain visible in evidence; configurations may differ.

Official / vendor headline scores; expand evidence for every record
Benchmark / evaluator	Headline score	Evidence
MathVision	96.8%unknown effort	The recorded result & source 96.8% · raw 96.8 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “MathVision(w/CI)”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). With a code interpreter, as the panel title says. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
Humanity's Last Exam (Multimodal)	30.1%unknown effort	The recorded result & source 30.1% · raw 30.12 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “Humanity's Last Exam-MM”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). The chart notes say HLE-MM uses the multimodal subset; tool use is not stated. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
ZEROBench (Ant Ling launch chart; tools and sampling unstated)	18%unknown effort	The recorded result & source 18% · raw 18 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “ZEROBench”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Solid series of the “ZEROBench / ZEROBench(Sub)” panel; tool use and sampling are not stated. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
ZEROBench(Sub) (Ant Ling launch chart; tools and sampling unstated)	43.7%unknown effort	The recorded result & source 43.7% · raw 43.71 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “ZEROBench(Sub)”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Hatched series of the “ZEROBench / ZEROBench(Sub)” panel; tool use and sampling are not stated. The post text states the same value. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
BabyVision	69.9%unknown effort	The recorded result & source 69.9% · raw 69.85 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “BabyVision”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Tool use is not stated (MathVision is the only panel labelled w/CI). Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
AntBench-Medical Comprehensiveness (Ant Ling internal)	89.5%unknown effort	The recorded result & source 89.5% · raw 89.52 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “AntBench-Medical Comprehensiveness”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Solid series of the “AntBench-Medical” panel, printed as a percentage. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
AntBench-Medical Correctness (Ant Ling internal)	99.2%unknown effort	The recorded result & source 99.2% · raw 99.22 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “AntBench-Medical Correctness”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Hatched series of the “AntBench-Medical” panel, printed as a percentage. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
OmniDocBench 1.5	92.2%unknown effort	The recorded result & source 92.2% · raw 92.21 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “OmniDocBench1.5”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Overall score as charted; the y-axis starts at 70. The post text states the same value. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
CharXiv (RQ split)	85.2%unknown effort	The recorded result & source 85.2% · raw 85.15 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “CharXiv_RQ”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Reasoning-question split; no tools are named. Kimi K3’s unstarred 84.80 equals Moonshot’s own no-tools value. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
GDP.PDF (vendor reports)	22.3%unknown effort	The recorded result & source 22.3% · raw 22.25 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “GDP.pdf”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Scoring rule not stated; the unstarred peer values equal the all-pass values other vendors cite. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
MedXpertQA MM	79.9%unknown effort	The recorded result & source 79.9% · raw 79.93 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “MedXpertQA-MM”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Multimodal subset; the y-axis starts at 50. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
Vision2Web	85.9%unknown effort	The recorded result & source 85.9% · raw 85.86 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “Vision2Web”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Harness and judge are not stated. The post text states the same value. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
AndroidWorld	85.6%unknown effort	The recorded result & source 85.6% · raw 85.63 % Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “AndroidWorld”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Harness is not stated; the y-axis starts at 50. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379
BlenderBench (Ant Ling launch chart)	67.8 scoreunknown effort	The recorded result & source 67.8 score · raw 67.77 score Headline · unknown effort · Own vendor Source/record date: 2026-11-05 Ant Ling’s Ling-3.2-flash launch post (5 November 2026), chart image panel “BlenderBench”, bar “Ling-3.2-flash”; value read from the image (https://pbs.twimg.com/media/HWYqPHCbUAAyVQb.jpg?name=orig). Tasks, harness and scoring are not stated. Reasoning effort and evaluation dates are not published. https://x.com/AntLingAGI/status/2118428246664761379

Independent evaluators

Evaluator harnesses are distinct from vendor measurements. Missing coverage is not a failed test.

No independent evaluator results are recorded for this model.

Read how we select and source scores or the comparison guide.