DeepSeek releases updated V4.1-Flash model it says completes tasks faster
SiliconANGLE Duncan Riley ● Covered by 3 sources
DeepSeek updated its V4.1-Flash model and says it runs tasks faster. It’s a small tweak, but the company is leaning hard on price and speed.
Based on reporting by SiliconANGLE, Duncan Riley — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
DeepSeek has pushed out a fresh version of its V4.1-Flash model, a little more than five weeks after the original arrived as the smallest model in its new family. The new build is called V4.1-Flash-1017, and DeepSeek’s own model card describes it as a minor update that cuts task time and shortens outputs. The company didn’t say by how much.
That restraint matters because the first V4.1-Flash launch was all about cost. In September, DeepSeek said off-peak API access would start at 15 cents per million uncached input tokens. After that release, Bloomberg reported that shares of MiniMax Group and Alibaba fell in Hong Kong.
The updated model looks a bit better in DeepSeek’s own testing. It solved 77% of tasks on the DeepSWE v1.1 software engineering test, up from 74.2% for the September version. Anthropic’s Claude Opus 5 is listed at 74% in the same table, though DeepSeek still says Anthropic’s model leads on seven of the 11 benchmarks it published. On Humanity’s Last Exam, the gap is still almost 14 points.
DeepSeek says it tuned the update for agent software, including its own Harness tool and Tencent’s CodeBuddy coding assistant and WorkBuddy desktop agent. WorkBuddy has already supported V4.1-Flash since the model first launched on Sept. 10. And DeepSeek also opened its in-house search engine to outside developers through a Search API and a Fetch API that turns a single web page into Markdown. Both are priced at 75 cents per 1,000 successful requests, with calls through DeepSeek Harness, Claude Code or Codex free until Nov. 1.
The release follows a nearly five-hour API outage on Oct. 15, which Yicai Global called one of DeepSeek’s worst this year. DeepSeek apologized, handed out credits based on what users had spent the week before, and said it would remove single points of dependency so the failure doesn’t happen again. The updated model is live in its web and mobile apps now, and API customers get it through the existing deepseek-flash model ID at the same prices.
My take — AI-written commentary, not fact-checked reporting
DeepSeek is doing the smartest thing in open AI right now: shipping smaller, cheaper, and just-good-enough models while everyone else talks about frontier glory. The outage was the reminder nobody asked for — cheap only matters if the plumbing holds. The industry keeps pretending scale is the whole story; users mostly care whether the thing works and bills them less.
Read more about this at: SiliconANGLE
Related stories
DeepSeek-V4-Flash Outshines Pro, The Biggest GitHub Crawl Yet, Engineering System Prompts for Safer Code
The Batch ·
43
DeepSeek Upgrades DeepSeek-V4-Flash-0731 with Major Agentic and Coding Gains
MarkTechPost · 2 months ago ·
55