This is even more insane, DeepSeek-v4.1-flash build me a complete game in one-shot attempt.
Not with any harness, not using any external assets, no tool calls, just single turn, one prompt in and html output. Model is DeepSeek-v4.1-Flash-1017 (deepseek-flash is the model name
DeepSeek open-sourced DeepInfer, the engine behind its API.
Its receipt: $38 of revenue per GPU-hour (off-peak). In 2025 the same math gave ~$13 on H800s. The catch: 64 + 64 Blackwell-class GPUs, and no cache store in the repo.
DeepSeek DeepInfer is the engine behind DeepSeek's API, open-sourced Oct 20, 2026.
DeepSeek says one GPU earns $38/hour off-peak. The tricks: split prefill/decode, decode across 64 GPUs, 96% cache hits (reported). The catch: the cache isn't in the repo.
You have reached the end of the archive
All of deepseek