DeepSeek-V4.1-Flash New Version Arrives on Volcano Ark: Zero-Quality-Loss 'Economy Edition' Launches at 20% Off

DeepSeek-V4.1-Flash-1017 is now live on Volcano Ark. Compared with the version first released on September 10, the new model posts significant gains over the previous version on Terminal-Bench 4.0, DeepSWE v1.1, ProgramBench and several other benchmarks. Task runtimes and output lengths have also decreased noticeably, reaching a new performance-cost Pareto optimum on multiple mainstream harnesses. With reinforcement learning based on self-feedback, the new model is significantly stronger at 'wish-style' programming tasks that build complete projects from scratch, such as game development and 3D modeling.

Alongside the standard version, Volcano Ark is also launching a DeepSeek-V4.1-Flash Economy Edition, available at 20% off during both idle and busy hours. The two-version offering addresses enterprises' differing requirements for performance, latency, and cost across business scenarios. Whether the use case is an interactive real-time application or large-scale background and batch processing, enterprises can call DeepSeek-V4.1-Flash at better value and accelerate AI application deployment.

Standard Edition: Consistent with Official Performance, with Extra Ark Capabilities

Volcano Ark's DeepSeek-V4.1-Flash-1017 stays consistent with the official version, so enterprises can call the full-capability model with zero barrier to entry, without deploying or maintaining complex architecture themselves.

On top of the original capabilities, Volcano Ark also provides enhanced features such as video understanding, explicit caching, JSON Schema, prefix continuation, role=developer, and connects seamlessly with the Volcano Engine cloud and AI ecosystem, flexibly invoking web search, MCP, and other tools to help enterprises build intelligent, automated applications more efficiently.

Try it on Volcano Ark: https://szacq.cn/juGUn/

Economy Edition: Identical Quality, 20% Off in Both Idle and Busy Hours

Building on the standard version, Volcano Ark is also launching the DeepSeek-V4.1-Flash Economy Edition.

Try it on Volcano Ark: https://szacq.cn/jhR8D/

The Economy Edition uses the same model weights as the standard version, so capabilities including agent, code, reasoning, multimodal, and 1M context are completely identical, with zero loss in model quality. For pricing, the Economy Edition is billed by time window at 20% off the standard version's published list price for the same period; calling it in any time window saves 20% compared with the standard version, supporting larger-scale model calls at lower cost.

Apart from higher response latency compared with the standard version, the Economy Edition remains consistent in quality and capabilities. It is especially suitable for the following scenarios:

• Batch data processing
• Background agent execution
• Training data synthesis
• Model evaluation batch runs
• Long-document understanding

Existing workloads can migrate quickly by simply replacing the Model ID (deepseek-v4-1-flash-economy-261017).

Two Versions, Flexible Choice: Good Models That Are Affordable and Usable

The standard version focuses on better latency and real-time experience; the Economy Edition cuts large-scale calling costs significantly with zero quality loss and 20% off in both idle and busy hours. With the two-version combination, Volcano Ark lets enterprises configure model resources flexibly according to business priorities, maintaining DeepSeek-V4.1-Flash's full capabilities while better balancing performance, latency, and cost, and supporting more AI applications moving from validation to large-scale deployment.

Agent Plan Also Upgraded to the 1017 Version

Both the personal and enterprise editions of Volcano Ark's Agent Plan have upgraded DeepSeek-V4.1-flash to the 1017 version, offering standard and economy options for developers to choose from.

The Agent Plan personal edition's limited-time usage-doubling promotion is underway. During the promotion, whether standard or economy, the DeepSeek-V4.1-Flash-1017 deduction coefficient is reduced by another 50%, doubling usage without an additional price increase. The promotion ends 12.30.

Volcano Ark is a one-stop large-model service platform under Volcano Engine. With the model at its core, it relies on four pillars: professional algorithm services, high-concurrency compute guarantees, a developer-friendly experience, and full-lifecycle security and trust, safeguarding large-scale model service deployment. The platform has served 1.1 million+ enterprises and individuals, covering 100+ industries, bringing large models deep into every sector.