Aug 14 2026

Alibaba releases Qwen3.8-27B under Apache 2.0, a multimodal 27B with 262K context built for 24GB consumer GPUs, while the 2.4T flagship keeps a custom license.

OPENMODEL

Original reporting

No source was cited for this one in the issue that logged it. The link above searches for it instead of guessing at a URL.

Native text, image and video input; hybrid gated attention plus DeltaNet; Alibaba's own scores of 61.7 and 84.3 OSWorld would make it the strongest single-GPU local model. The license split, frontier gated and deployable tier free, is the strategy.

Named in this event

Alibaba

Open calls that touch this

Alibaba ships the Qwen3.8-Max weights by August 16th · 65%, resolves August 16th 2026
A second flagship API adopts time-of-day pricing · 55%, resolves February 17th 2027

Issues that mention Alibaba

Around the same time

Aug 14 2026MONEYINFRA

SpaceX closes its $60B acquisition of Cursor, the largest venture-backed startup acquisition on record, folding the coding tool into SpaceXAI alongside xAI.

search for coverage

Aug 14 2026ADOPTINFRA

Uber and Pony.ai will deploy 2,000+ robotaxis across Europe, expanding beyond their Zagreb launch to four more cities plus the Middle East.

search for coverage

Aug 14 2026MONEY

Anthropic posts its first profitable quarter, with preliminary Q2 revenue above $11.5B, up 14-fold year over year, ahead of an expected October listing.

search for coverage

Aug 14 2026POLICYADOPT

Apple trained its own China model with Alibaba's help, the first foreign proprietary model ever cleared by Beijing's algorithm registry.

search for coverage

Aug 14 2026PEOPLESAFETY

OpenAI's pre-IPO leadership purge comes into view: COO, CRO, ethics chief, safety head and chief futurist all gone in weeks as power consolidates under Greg Brockman.

search for coverage

Aug 14 2026SAFETYMODEL

Anthropic raises its misalignment risk rating to low and discloses unreleased Model 2

search for coverage