DeepSeek V4.1 Flash 1017 tested: up to 45% fewer thinking tokens, #1 open model on Arena WebDev
Prompt Engineering · youtube · 2026-10-19
The Prompt Engineering channel tested DeepSeek's 'minor' V4.1 Flash 1017 update and found more than expected: every benchmark in DeepSeek's table improved and it's now the #1 open model on Arena's WebDev leaderboard. The bigger story is efficiency — with identical prompts and settings, 1017 used roughly a third fewer thinking tokens at high effort and 45% fewer at max effort, addressing Flash's token-hungry reputation (cheap per token but not per task). Demos include an SDF dragon in one shader, 3D world generation, galaxy collision on a DGX Spark, robot walking, floor-plan-to-3D walkthrough, double pendulum vs Gemini 4 Argon, and screenshot-based website rebuilds.
Related event: DeepSeek V4.1-Flash-1017: 'Minor Update' Tops Open-Weight Leaderboards(12 posts)→
More from Models
- lintool's Gaggle Matches SFT'd Qwen With Just a Handful of GPUs on nanochat Backbone — lintool · 2026-10-20
- Report: Liang Wenfeng planned independent DeepSeek compute chips before Huawei persuaded him otherwise — teortaxesTex · 2026-10-20
- Smallest AI's full voice stack hits 26.8% completion at $0.048/call on 300 hard calls — auto_grad_ · 2026-10-20
- NVIDIA's Dory1.0-4B Looped Model Lets You Pick Inference Depth at Runtime — lmoroney · 2026-10-20
- OpenAI dropped 722 AI-generated math papers in a day; can humans read them? — songyoupeng · 2026-10-20
- JevBench adds 34 decision models in four days; Nimble-v4-max price cut to $0.04/1M tokens — airesearch12 · 2026-10-20