Breaking

Alibaba Unveils Qwen3.8-Max Preview, a 2.4-Trillion-Parameter Model It Ranks Behind Only Anthropic's Fable 5

July 19, 2026 at 07:16 EDT

Foundation Models
Open Source

Alibaba's Qwen team has released a preview of Qwen3.8-Max, a 2.4-trillion-parameter flagship model the company describes as comparable to leading frontier AI systems and second only to Anthropic's Fable 5. The Qwen3.8-Max-Preview is already available through Alibaba's Token Plan, Qoder and QoderWork, with an open-weight version promised in the near future.

Alibaba Qwen · Model Preview

Qwen3.8-Max: a 2.4-trillion-parameter bid for the frontier

Alibaba's largest model yet is out in preview — framed as rivaling top-tier AI and trailing only Anthropic's Fable 5, with an open-weight release promised to follow.

2.4T

parameters — the largest Qwen model to date

~2×

the scale of Sept 2025's 1T+ Qwen3-Max preview

rank claimed — second only to Fable 5

Doubling the scale

Parameter count vs. the earlier Qwen3-Max preview — each block ≈ 0.4T parameters

~1T+

Qwen3-Max
(Sept 2025)

2.4T

Qwen3.8-Max
(preview)

What Alibaba emphasizes

A text-centric instruct model pairing high-end capability with an eventual open-weight release.

Coding gains Long-context handling Data analysis Open weights (promised)

Developer optimism

Strong interest in a near-frontier model arriving as open weights — potentially one of the largest openly available models, sharply expanding who can build on and fine-tune it.

Caution needed

The Fable 5 comparison needs independent verification — no published benchmark scores, pricing, or concrete real-world use cases at the preview stage.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.