DeepSeek V4.1 Flash hands-on tests and comparisons

An updated comparison video puts Kimi K3.1 Pro at 46.0 versus DeepSeek V4.1 Flash 1017 at 32.5 under best settings.

The post cites Artificial Analysis best-setting scores of 46.0 for Moonshot Kimi K3.1 Pro versus 32.5 for DeepSeek V4.1 Flash 1017, with Moonshot ahead.

A user reports that V4.1 Flash 1017 optimized its own inference on H200s, yielding a 26% speedup in a single day, with a screenshot.

Using a $2 Blender render as a hook, the post notes V4.1 Flash ships open weights under MIT at the same low price with better reasoning, coding and agents, and asks whether open weights are now the default.

After the October DGX Spark system update, usable memory rises by 2GB and a dual-machine DeepSeek V4.1 Flash recipe runs about 5% faster in testing.

A user complains GPT has become slow and says running DeepSeek V4.1 Flash 1017 on own hardware gives thousands of tokens per second, which they can no longer live without.

The author argues the V4.1 Flash 1017 build leads OpenRouter's paid developer traffic on price and token efficiency, not overall capability.

OpenRouter's weekly ranking puts DeepSeek V4.1-Flash-1017 first at 33.6 trillion tokens, ahead of V4.1 Flash, MiniMax M3.1 Flash, GLM 5.3 Flash and Hunyuan Hy4 preview, with Chinese models leading for 26 straight weeks.

The author deployed DeepSeek V4.1 Flash 1017 on four DGX Sparks with two self-made recipes, SGLang and TensorFold, and the attached image shows the comparison results.

The author ran Cohere's prefill method 100 times on DeepSeek-V4.1 (1017) with zero supplied URLs; the model's self-generated hosts were mostly legitimate threat-intel sources, and 169.254.169.254 only appeared when explicitly asked to read cloud metadata.

A user says they have started using a locally deployed DeepSeek-V4.1-Flash-1017 for real work and find it very comfortable, without citing specific tasks or numbers.

A user recaps DeepSeek's releases from April to October: V4 preview, a permanent 75% price cut, DSpark and DeepSpec, V4-Flash GA, V4-pro-0813, V4.1-flash and DSec, ending with V4.1-flash-1017 and DeepInfer in October.