JD Yanxi 2
JD.com · Undisclosed (MoE, retail-optimised)
The retail specialist for product images and supply-chain documents
JD’s scenario model: product-image understanding, SKU attribute extraction and supply-chain documents are home ground, with e-commerce imagery and Chinese product text far more represented in training than at generalist vendors. Composite 78.0, ranked 28th, one of the models strictly dominated by MiMo-V2.5 in the 9-dimension analysis — but the model dominating it does not understand product images.
Last updated Oct 18, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- JD Yanxi 278.0
Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.
Public benchmarks
as reported, not converted| Benchmark / metric | Value | Note |
|---|---|---|
| 商品属性抽取 F1 | 93.7 | Vendor figure |
| 输出单价 | ¥6 | Official API |
Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.
Scenario mapping
five stress tests, one scale| Scenario | Score | Observation |
|---|---|---|
| Chinese receipt OCR | 85 · Good on document scenarios | |
| Chart data extraction | 76 · Standard charts | |
| 30-min video QA | 52 · No video modality | |
| Mobile GUI control | 72 · Not trained for GUI | |
| Product-image fine-grained recognition | 90 · Exclusive scenario corpora |
Strengths and limitations
Strengths
- Exclusive product-image and SKU corpora: fine-grained attribute extraction is home ground.
- Deep accumulation on supply-chain documents and Chinese e-commerce text.
- Complete toolchain for customer service and after-sales scenarios.
Limitations
- Strictly dominated by same-priced MiMo-V2.5 on all nine dimensions: leads on nothing.
- Video 55: no video modality support; long-horizon tasks absent.
- Ecosystem bound to JD’s stack: read cross-platform restrictions carefully.
Price and real cost
List price
¥6/M tokens
≈ ¥6 / 1M output tokens. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥3.6 (assuming input at 25% of output price).
Use your volumeAbout JD Yanxi 2
Is JD Yanxi 2 open source?
Licence: Closed commercial. No public weights; access is via vendor API or an on-prem licence.
How much does JD Yanxi 2 cost?
≈ ¥6 / 1M output tokens. It ranks #28 with a composite of 78.0. Real bills depend on resolution, tiling, caching and volume discounts.
What is JD Yanxi 2 best for?
Product tagging and SKU attribute extraction; Supply-chain document structuring; E-commerce customer service and after-sales QA。
Why does JD Yanxi 2 score 78.0?
The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is Chinese context & culture; its weakest is Video & long-horizon reasoning.
Personal ranking · Not an official score
Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.