DeepSeek Launches V4 AI Models in Mid-July and Doubles Prices During Peak Hours
DeepSeek announced Tuesday that it will launch its V4 large language model series for general access in mid-July, while simultaneously introducing a pricing structure that doubles token costs during peak business hours in Beijing. The update email, distributed to early users, states the release will bring "functional optimizations and performance improvements" following the model's initial preview phase earlier this year.
The launch coincides with the publication of DSpark, a speculative decoding framework DeepSeek says boosts AI inference speeds by up to 85% and has been tested on open models like Gemma and Qwen. Despite the rate doubling during Beijing's core work hours, DeepSeek's peak pricing of roughly $1.76 per million output tokens for the V4-Pro version remains significantly below major Western competitors, which charge around $15 per million tokens, keeping the firm at the forefront of an intense AI pricing competition.
From the sources (9 posts)
@askalphaxivRT @askalphaxiv: DeepSeek just published DSpark, a speculative decoding system that boosts live DeepSeek V4 serving throughput by 51% to 40…
@teortaxestexExtending DSpark to GDN-based models will be huge I hope Kimi team in K3 uses it (or some more advanced acceleration in this genre) from the get-go. This also speeds up RL…
@autotrustaiFirst model to support DSpark speculative decoding.@deepseek_ai DeepSeek-V4-Flash-DSpark-4E: 284B MoE → only 11B activated per token 18% faster inference, +3% MMLU-Pro accuracy 1M context. Fully open (MIT).
@teortaxestexDeepSeek V4 is coming. Mid-July. Yeah yeah you might think we had V4 for over 2 months already, but no, that was "preview of V4". They expect a lot of demand, and introduce a new mechanism: peak hour pricing. It doubles what we have now. T
@poezhao0605DeepSeek announced peak/off-peak pricing for V4, launching mid-July. Rates double during Beijing business hours (9-12, 14-18). The numbers: V4-Pro output at peak costs RMB 12 per million tokens, roughly $1.76. Anthropic's comparable Sonnet
@techmemeDeepSeek details DSpark, a speculative decoding framework for its V4 models, saying it speeds up AI inference by up to 85% and was tested on Gemma and Qwen (Ben Jiang / South China Morning Post) (Visit Techmeme dot com for the link and ful
@buildwithhassandeepseek V4 full release mid-july. adding peak hour pricing that doubles the cost. V4-Pro output: ~$0.84/M off-peak → ~$1.68/M peak V4-Flash output: ~$0.28/M off-peak → ~$0.56/M peak peak hours: 9:00-12:00 and 14:00-18:00 beijing time. s
@zyvoraxiaChinese DeepSeek users just received an email with big news: V4 official version is scheduled to launch in mid-July. At the same time, they’re introducing peak-valley pricing: during peak hours (Beijing time 9:00-12:00 & 14:00-18:00, tota
@lentils80Some users are getting this email from DeepSeek, informing them that DSV4 will leave Preview and enter "GA" (general access) in mid July "This version update will bring more functional optimizations and performance improvements" Also, the