DeepSeek's "Minor" Update Isn't Minor (V4.1 Flash 1017 - Fully Tested!)
DeepSeek calls V4.1 Flash 1017 a “minor update”, but it went up on every benchmark in their table, and it’s now the #1 open model on Arena’s WebDev leaderboard. The bigger story is tokens. DeepSeek has always been cheap per token, but its Flash models were token-hungry, so cheap per token never meant cheap per task. In my reruns, 1017 used about a third fewer thinking tokens at high effort and 45% fewer at max, with the same prompts and settings. 🔗 Links DeepSeek announcement: https://x.com/deepseek_ai/status/2111375640717033777 Model card (Hugging Face): https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017 DeepSeek API change log: https://api-docs.deepseek.com/updates Arena WebDev leaderboard: https://x.com/arena/status/2111894091168682086 Artificial Analysis (token usage): https://artificialanalysis.ai/models/deepseek-v4-1-flash Gemini 4 Argon on the same tasks: https://youtu.be/yksrd5uNxcw whryte, local speech-to-text for Mac: www.whryte.com Website: https://engineerprompt.ai/ RAG Beyond Basics Course: https://prompt-s-site.thinkific.com/courses/rag Let's Connect: 🦾 Discord: https://discord.com/invite/t4eYQRUcXB ☕ Buy me a Coffee: https://ko-fi.com/promptengineering |🔴 Patreon: https://www.patreon.com/PromptEngineering 💼Consulting: https://calendly.com/engineerprompt/consulting-call 📧 Business Contact: engineerprompt@gmail.com Become Member: http://tinyurl.com/y5h28s6h 💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off). 00:00 DeepSeek's "minor" update 01:10 What's new in 1017 02:02 #1 open model on Arena WebDev 02:53 Why DeepSeek Flash was token-hungry 05:04 SDF dragon in one shader 05:41 Three 3D worlds 06:11 Galaxy collision on a DGX Spark 07:05 Teaching a robot to walk 07:50 Image and video input 08:21 Floor plan to 3D walkthrough 08:51 Double pendulum: DeepSeek vs Argon 09:52 Website rebuilds from a screenshot 10:32 Arena vs copying a page 10:55 Dark mode and photos drawn in SVG 11:31 High vs max effort Credits: SDF dragon experiment by Keenan Crane, concept art by Inigo Quilez (CC0). Double-pendulum footage by F. Stollmeier & M. Schottdorf (CC BY-SA 4.0).
