Topic · 10 stories
10 stories tagged deepseek on AI Tamer, newest first.
Fangcun ran 34 evaluations on DeepSeek's newest Flash model. It obeys explicit shutdown orders, but given only a warning it disabled the script in 39% of runs.
Unit 42 says an attacker it calls BUGSTER used Hermes Agent, DeepSeek V4 and a Cursor setup with safety off to hit Indian universities and plan VoIP fraud.
In one unpublished Vals AI scenario, DeepSeek V4.1 Flash averaged 16.3 of 21 objectives to GPT-6 Sol's 15.7. On public CyberBench, Sol leads.
DeepInfer is Apache-2.0 and runs V4.1-Flash on NVIDIA SM100 or Huawei Ascend 950. This release is offline inference only, without KV cache or load balancing.
A FASCSA order excludes DeepSeek from civilian agency purchases and removes its products, plus third-party products developed using DeepSeek implementations.
A ByteDance Seed preprint says token offset can swing DeepSeek-V4 retrieval. The 40-point gap is the base checkpoint. Post-trained gaps are smaller.
DeepSeek published DeepSeek-V4.1-Flash-1017 under MIT on Hugging Face. Callers still use deepseek-flash, and DeepSeek says the price is unchanged.
DeepSeek's API guide documents Search and Fetch at 5 requests a second. The Chinese pages price each at 5 yuan per 1,000 successes; the English pages say $0.75.
River Inference is serving DeepSeek V4.1 Flash and GLM 5.3 Flash. Its public card is 20% under DeepSeek's off-peak rates and Z.ai's list prices. A second homepage table is much higher.
Bloomberg reports DeepSeek is close to raising at least 80 billion yuan, with Tencent and CATL among the largest commitments, ahead of an early 2027 IPO. Reuters said it could not verify the report.
