Kairos 3.0-4B
SenseTime · 4B dense
A 4B world model
SenseTime’s 4B world model open-sourced in June 2026 under Apache-2.0: video prediction 80, generation 78 and usable embodied scenarios, for a composite of 66.3. Documents 50, charts 45 and Chinese 75 show it is not a general assistant but a lightweight base for world-model and embodied research.
Last updated Oct 21, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- Kairos 3.0-4B66.3
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 |
|---|---|---|
| 参数规模 | 4B | Dense model |
| 许可证 | Apache-2.0 | Commercial use |
| 上下文 | 128K | Official spec |
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 |
|---|---|---|
| World-model prediction | 82 · A new capability axis here | |
| 30-min video QA | 80 · Works on dynamic scenes | |
| Embodied scenarios | 80 · Simulation and planning | |
| Chinese receipt OCR | 50 · Not a document model | |
| Chinese cultural context | 75 · Mostly research context |
Strengths and limitations
Strengths
- A 4B world model on one GPU keeps research and prototyping cheap.
- Video prediction 80 and generation 78 make dynamic-scene modelling viable.
- Apache-2.0 lets embodied and simulation projects modify it freely.
Limitations
- Documents 50 and charts 45: cannot carry image-text business work.
- Chinese 75 and agents 55: weak as a general assistant.
- Ecosystem 82: research-oriented, with thin production docs.
Price and real cost
List price
—N/A
Open weights; runs on one GPU. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥0.0 (assuming input at 25% of output price).
Use your volumeAbout Kairos 3.0-4B
Is Kairos 3.0-4B open source?
Licence: Open weights (Apache-2.0). Weights are available, but check the exact terms and filing requirements before commercial use.
How much does Kairos 3.0-4B cost?
Open weights; runs on one GPU. It ranks #54 with a composite of 66.3. Real bills depend on resolution, tiling, caching and volume discounts.
What is Kairos 3.0-4B best for?
World-model research; Video prediction and simulation; Embodied-intelligence prototypes。
Why does Kairos 3.0-4B score 66.3?
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 Visual perception & grounding; its weakest is Chart & document 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.