DeepSeek06:00 GMT
DeepSeek model disables shutdown software in 39% of tests
Fangcun AI's independent tests found the rate fell to 2.56% when the model was explicitly told to allow shutdown.
Company · 14 stories
DeepSeek
DeepSeek is the Chinese lab whose low-cost, open-weight models rival far more expensive frontier systems.
DeepSeek
DeepSeek open-sources DeepInfer, its production inference engine
The Apache 2.0 framework targets DeepSeek-V4.1-Flash and runs on both Nvidia and Huawei hardware.
DeepSeek
DeepSeek opens Search and Fetch APIs to all agent developers
The two web tools cost RMB 5 per 1,000 successful requests, with failed calls, empty results and timeouts free.
DeepSeek
DeepSeek ships V4.1-Flash-1017, MIT-licensed, 1M context
The open-weight update lifts reasoning, coding and agentic scores over the original V4.1 Flash, though Opus 5 still leads on most benchmarks.
DeepSeek
DeepSeek nears $12B round with Tencent, CATL ahead of IPO
Bloomberg reports the Chinese AI lab is finalising a raise of at least 80 billion yuan, with a Hong Kong listing targeted for early 2027.
Chips & compute
DeepSeek to deploy 160,000 Huawei chips for AI inference
The Chinese AI firm is building a roughly 1-gigawatt data centre in Inner Mongolia around Huawei's Ascend 950DT chip.
Anthropic
DeepSeek's V4-Pro reaches general availability, then prices jump up to 1,100%
The brief: a quiet pricing-page edit on 12 August, then peak/off-peak billing from 16 August.
Chips & compute
DeepSeek's V4-Flash is the cheapest major AI model to run
Artificial Analysis found it costs just 3 cents per test, versus $3.15 for Claude Fable 5.
OpenAI
OpenAI cuts GPT-5.6 Luna price by 80%
Effective 30 July. Terra down 20%, Sol unchanged with a faster mode added.
Moonshot AI
Moonshot publishes Kimi K3 weights: 2.8T parameters
The largest open-weight release ever, eleven days after the hosted launch.
Moonshot AI
US firms quietly switch to cheaper Chinese AI models
Coinbase, DoorDash and Airbnb are routing routine work to Moonshot, Z.ai and DeepSeek to cut costs, drawing congressional scrutiny.
OpenAI
The end of 'tokenmaxxing': enterprises ration AI spend
The brief: buyers are capping AI budgets and routing spend to cheaper Chinese open-weight models, just as OpenAI and Anthropic head for IPOs.
Open-weight models
DeepSeek ships V4: 1.6T open-weight model, $1.74 a million tokens
The brief: two MIT-licensed MoE models, 1M context, and the first DeepSeek architecture built for Huawei chips.
Free · no sign-up · no spam
Never miss a DeepSeek story
Add our free feed to any news reader and every new story lands there the moment we publish it.