84634E1A607A

Testing version name: deepseek-v4.1-flash-expires-on-0910. DeepSeek is reportedly planning to officially release its V4.1 Flash model around September 10, 2026 (Beijing Time). According to internal and external testing, said DeepSeek, V4.1 Flash has outperformed V4 Pro across key metrics, including performance, cost, generation speed, and overall completion time. After V4.1 Flash is officially launched and before V4.1 Pro becomes available, requests to V4 Pro will be routed to V4.1 Flash instead and billed at the V4.1 Flash rate. DeepSeek is also asking users to report any issues identified when comparing V4 Pro and V4.1 Flash.

The official Chinese release post has an interesting lead-in:

> 在一些前沿模型突然变得不可用的时刻，我们选择相信另一条路 from https://mp.weixin.qq.com/s/LDrbtLM0wiCTJorvd5GY9w

In the moment some frontier model suddenly becomes inaccessible (You, Fable), we choose to believe in another way.

Happy to see the open-weight model being ambitious to compete with the frontier closed-weight ones, and hope (well, maybe wish?) the smaller-sized ones will one day fit on truly consumer-grade GPUs to run locally, while still competent enough to do heavy agentic work. The future of AI should be open enough that people can build on it, not rent access to it.

Yep, but this is server-side setting. Were I a sniffer, I would set this to 10000 and now I can correlate keys.

Overall good work. I'd request an `-i` command-line parameter to specify the interface to scan (and I'd prefer ALL params being able to be read from command line params). I think it just performs a full scan initially on my laptop, following scans either didn't success or didn't involve TCP connect scan (I don't see ARP requests after the initial scan).

/56 DHCPv6 Prefix Delegation, and NAT44444 for IPv4 (192.168.1.x/24 for router, 192.168.100.x/24 for dish, 100.64.x.x for CGNAT, 172.16.x.x for Ground Station inner net, finally you are on the Internet).