Alibaba's DAMO RADAR Beat 23 of 26 Radiologists on CT Scans
A.I. / news
Alibaba's DAMO RADAR Beat 23 of 26 Radiologists on CT Scans
The open-sourced model scored an average 0.913 AUC across 146 abdominal conditions in a 40,000-exam study, but its GitHub and Hugging Face pages disagree on whether anyone can use it commercially.
Alibaba's DAMO Academy said Sept. 18 it has open-sourced DAMO RADAR, a vision-language model that reads contrast-enhanced abdominal CT scans and flags nearly 150 conditions, including several cancers, in a single pass. The research, published in Science as Zhang et al., 393(6817) per the project's own citation record, reported an average area under the curve of 0.913 across 146 clinical findings spanning 18 organs, measured against roughly 40,000 real-world examinations, according to South China Morning Post.
An AUC, a standard measure of a diagnostic test's ability to separate true cases from false ones, tops out at 1.0 for a perfect classifier; DAMO Academy's institute description called the model "the world's first expert-level generalist medical imaging model" and said the training approach could extend to other imaging types beyond CT.
How it compares to the radiologists who tested it
In a reader study pitting RADAR against 26 radiologists from different hospitals, the model's average diagnostic accuracy beat 23 of them, Yicai Global reported. Zhang Ling, a senior algorithm expert at DAMO Academy, said the model raised radiologists' disease-detection sensitivity by 10 percentage points and cut diagnostic time by 30 percent when used as an assistive tool rather than a standalone reader. The model was trained on more than 400,000 contrast-enhanced CT examinations paired with 15 million anatomy-aware image-text pairs drawn directly from clinical reports, according to the project's GitHub page, without manual annotation of the underlying images.
- Outperformed by RADAR23 radiologists
- Not outperformed3 radiologists
Source: Yicai Global, accessed 2026-09-19
Two licenses for the same weights
DAMO Academy's GitHub repository, created July 3, 2026 and carrying 191 stars and 14 forks as of publication, lists the code under an Apache 2.0 license, which permits commercial use without restriction. The model weights on Hugging Face, hosted separately at the project's own page, carry a CC BY-NC-SA 4.0 license instead, a Creative Commons term that explicitly bars commercial use and requires any derivative to carry the same restriction. Neither page acknowledges the mismatch, and it is not clear which license actually governs a hospital or vendor that wants to deploy the weights in a paid product rather than research.
| Location | License | Commercial use |
|---|---|---|
| GitHub repository | Apache 2.0 | Permitted |
| Hugging Face model weights | CC BY-NC-SA 4.0 | Barred |
What DAMO hasn't said
Neither DAMO Academy's own materials nor the sources cited above mention a regulatory clearance in China or elsewhere, a deployment at any named hospital, or a timeline for one. The Science paper establishes retrospective, reader-study performance rather than a clinical trial with a treatment outcome, and Alibaba has not said whether it plans to submit RADAR for medical-device approval before any hospital uses it to make a diagnosis rather than double-check one. DAMO Academy, founded in 2017 as Alibaba Group's research arm, has increasingly split its open-source releases between commercially licensed developer tools, like the IntelliJ plugin it shipped for its Open Code Review product this month, and research-only releases like this one; RADAR's contradictory licensing suggests that split was not fully thought through before publication.
RADAR is Alibaba's second major open-weight release this month, after Alibaba Cloud's Qwen3.8-27B language model, though the two come from different divisions of the company and carry no shared licensing terms. The next thing to watch is whether DAMO Academy corrects the license mismatch, or clarifies which one actually applies to anyone trying to build on the weights rather than just read the paper.
Sources
- alibaba-damo-academy/damo-radar, Alibaba DAMO Academy (GitHub)
- RADAR: An Expert-Level Generalist AI for Abdominal CT Diagnosis, Hugging Face
- Alibaba open-sources medical AI model that can detect cancer and nearly 150 conditions, South China Morning Post
- Alibaba's Damo Academy Debuts Generalist AI for Nearly 150 Abdominal Conditions, Yicai Global
Sources
More in A.I.
- 01GPT-Rosalind-6 Reuses GPT-6 Astra's Weights, Cuts Biology Refusals to 9.4% and Stays Behind a GateOpenAI's Oct. 30 system card addendum says no training was added, so the difference from Astra is deployment controls and who is allowed in.
- 02Architect Says Claude Opus 5.5 Drew a House Set Good Enough for a Permit After Two Redline RoundsStephen Gallagher's informal test of 13 models gave Opus 5.5 the most human feedback, and the drawings still missed a required second exit.
- 0350 GPT Agents in Minecraft Built a Ration Ledger, and 34 Died Doing ItY Combinator's Labs group and Olam Labs ran GPT-6.1 Sol and Claude Opus 5.5 swarms through Minecraft and Mindustry. GPT won both head-to-head games, but each result comes from one harness and one setup.
- 04China's Announcement No. 48 Puts 23 AI Technologies on a Draft Export ListExporters of listed technology must already register with MOFCOM. Comments on the draft catalogue close on November 15, and the sources disagree on whether the thresholds are published.