Alibaba's new Qwen model for voice and video costs more than 98 percent less per hour of audio than the one before
The AI stack on 20 September 2026. 2 moves, 1 with a primary source.
Alibaba's new Qwen model for voice and video costs more than 98 percent less per hour of audio than the one before
· L4 Models · >98% lower per hour of audio input; >93% lower for audio and video; 1M-token context · Confirmed
Alibaba's Qwen team released Qwen3.8-Omni-Flash on 20 September, a model that takes in text, audio and video and can carry out tasks with them, such as editing video or summarising long meetings. Alibaba says the API price per hour of audio input is more than 98 percent lower than for its Qwen3.5-Omni-Plus, and more than 93 percent lower per hour of audio and video input. It notes these hourly prices are its own estimates, worked out as 30 times the cost of two minutes of material. The model can take in up to 1 million tokens, the pieces of text and media it reads, at once.
Thailand proposes treating data centres of 100 megawatts or more as 'hyperscale', with new rules expected by mid-October
· L1 Power and land · 100 MW threshold proposed; rules expected by mid-October · Reported
Thailand's National Economic and Social Development Council is proposing that data centres using 100 megawatts of power or more be classed as 'hyperscale', according to The Nation, with new rules on size, electricity use and location expected to take effect by mid-October. The threshold is not final and still goes to a government policy committee. For people in Thailand, the planned rules include a requirement to avoid harming households.