DeepSeek V4.1 Flash updated to DeepSeek-V4.1-Flash-1017, and the in-house search API is opened
DeepSeek-V4.1-Flash-1017 更新(同页条目:开放自研搜索 API)
AI summary
DeepSeek V4.1 Flash has been upgraded to DeepSeek-V4.1-Flash-1017, with the official web interface and app updated simultaneously. The API calling method and pricing remain unchanged, still using deepseek-flash. Compared with the previous DeepSeek V4.1, the new model improves on benchmarks such as HLE (w/o tools) 42.6 and DeepSWE v1.1 77.0, and is said to reach a new performance-cost Pareto optimum under multiple Harnesses. The same page also opens the in-house search API.
Read original