A developer who has used DeepSeek 4.1 Flash heavily for about a month across a dozen projects says it is highly capable and orders of magnitude cheaper than frontier models. Mid-session, they often cannot tell whether they are using DeepSeek or Opus, and they report no noticeable difference in conversations, work quality, or speed. They argue the model behaves like a frontier model, and they question why frontier labs are not more alarmed, noting that Chinese distilled models may be only a month or two behind Anthropic and OpenAI while handling the same workloads.
The author says today's models are good enough for high-quality unattended tasks, so chasing the latest releases is often unnecessary. With a $10 monthly subscription, DeepSeek is effectively unlimited, which has changed their development approach. They use it for mindless tasks, exploratory testing, complex planning, and research, and occasionally bring in Opus 5.5 for a final code review before having DeepSeek apply the fixes. Most sessions cost well under $1.
The author attributes this affordability largely to DeepSeek's cache optimizations, which reportedly shrank the KV cache by roughly 437 times compared with its V1 model. Holding that cache in GPU memory is one of the biggest costs of long coding sessions, and the author suggests the approach may also use less water and electricity than Claude.
The author still holds frontier subscriptions through work and says the goal is long-term planning, sustainability, and broader access to high intelligence. They see the tech industry's focus on paying top dollar as misguided. On self-hosting, they argue it is not economical for saving money but may become worthwhile for privacy as cache optimizations allow the system to run locally.
Source: Hacker News · Summarized by HeadlinesBriefing