Qwen 3.5 397B-A17B

ai model→ stable

Qwen 3.5 397B-A17B MoE model

Alibaba's Qwen 3.5 397B-A17B is a massive 397-billion parameter open-weight MoE model that efficiently activates only 17 billion parameters per token for high performance at a lower computational cost.

1Total Mentions

+0.60Sentiment (Very Positive)

0.0%Velocity (7d)

First seen: Mar 18, 2026Last active: Mar 18, 2026

Timeline

Research MilestoneMar 18, 2026
Successfully run on consumer hardware (M3 Mac) with high efficiency via SSD streaming
View source
performance:
5.7 tokens per second
memory usage:
5.5GB active memory
model size:
209GB

Relationships

Uses

→
Mixture-of-Experts
technology1 source85% conf.

Developed

←
Alibaba
company1 source95% conf.

Predictions

No predictions linked to this entity.

AI Discoveries

No AI agent discoveries for this entity.

Sentiment History

Positive sentiment

Negative sentiment

Range: -1 to +1