Qwen3-Omni-30B
Alibaba · 30B-A3B (MoE)
The open option that genuinely unifies audio
Image, audio, video and text through one interface, with low 30B-A3B inference cost. The most realistic open starting point for omni prototypes, though 32K context forces chunking.
Last updated Oct 18, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- Qwen3-Omni-30B88.9
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 |
|---|---|---|
| 模态覆盖 | 4 | Text/image/audio/video |
| 激活参数 | 3B | 30B total MoE |
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 | 89 · Fine on clean receipts | |
| Chart data extraction | 86 · Occasional misreads on complex charts | |
| 30-min video QA | 87 · Needs chunking; average timing | |
| Mobile GUI control | 82 · Not a strength | |
| Chinese cultural context | 90 · Solid Qwen Chinese base |
Strengths and limitations
Strengths
- Unified text/image/audio/video input, less glue code.
- Apache-2.0 with few commercial restrictions.
- Small active parameter count keeps inference affordable.
Limitations
- 32K context hurts long video and long docs.
- Chinese dialect ASR is average.
- Complex chart reasoning trails flagship VLMs.
Price and real cost
List price
¥4/M tokens
Open; API ≈ ¥4 / 1M output. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥2.4 (assuming input at 25% of output price).
Use your volumeAbout Qwen3-Omni-30B
Is Qwen3-Omni-30B open source?
Licence: Apache-2.0. Weights are available, but check the exact terms and filing requirements before commercial use.
How much does Qwen3-Omni-30B cost?
Open; API ≈ ¥4 / 1M output. It ranks #10 with a composite of 88.9. Real bills depend on resolution, tiling, caching and volume discounts.
What is Qwen3-Omni-30B best for?
Joint speech and frame analysis; Open omni-modal prototypes; Cost-sensitive batch multimodal tasks。
Why does Qwen3-Omni-30B score 88.9?
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 Document parsing & Chinese OCR; its weakest is GUI agents & tool use.
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.