Summary
DeepSeek has released V4.1 Flash 1017, an upgrade to its low-cost Flash model that arrived about five weeks after the original V4.1 Flash. This video from TheAIGRID breaks down what the release means for the AI market.
The model is open weights under the MIT license, handles text and images, and offers a context window of around 1 million tokens. DeepSeek’s API prices it at $0.30 per million input tokens and $1.20 per million output tokens, far below Claude Opus 5 at $5 and $25 or GPT 5.6 Sol at $4 and $20. In DeepSeek’s own benchmark table, it scores 77% on its software engineering test, ahead of Opus 5 at 74, and leads on Automation Bench and Apex Bench. It trails Opus 5 on Humanity’s Last Exam and Terminal Bench 4.0.
The video also covers community tests, including a Tesla Roadster driving-scene build that cost $0.66 on DeepSeek versus about $110 on Opus 5.5, and a comparison with MiniMax M3.1 Flash. It closes with the political fallout, including a US order to remove DeepSeek from civilian federal computers and a report of a hacker using the model against more than 100 companies.
📺 Source: TheAIGRID · Published October 28, 2026
🏷️ Format: News Analysis
