# Baseten — How to make flash models even faster: optimizing DeepSeek V4.1 Flash 1017

- Company: Baseten (baseten.co)
- Announced: 2026-11-04
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.baseten.co/blog/how-to-make-flash-models-even-faster-optimizing-deepseek-v41-flash-1017-fast
- Record: https://forck.live/items/34078-how-to-make-flash-models-even-faster-optimizing-deepseek-v4-1-flash-1017
- Subject: Baseten
- Models affected: DeepSeek V4.1 Flash 1017

Baseten optimized 80+ kernels and bottlenecks for DeepSeek V4.1 Flash 1017 to produce a Fast variant, which it says reaches 991 output tokens/s on Artificial Analysis and the lowest time per task on the Pareto frontier. The company says flash models now dominate the low-cost Pareto frontier, and it cites DeepSeek V4.1 Flash 1017 at native precision as a 552B-parameter MoE with 8B active during prefill and 16B active during decode that scores 32 on the new Artificial Analysis V5.0 Index, the same as GLM-5.3, at roughly a third the cost and 4x the speed of the larger model. Baseten describes three hardware-level techniques behind the work: kernel fusion, Programmatic Dependent Launch and host loop minimization, and says it plans to release open-source work around each soon. One example is an improved fused router kernel that decreases decode time by 2.7ms, and the team says it improved the open-source megakernel for mHC in DeepInfer for lower latency at batch size 1 and 30% more speed. In Cline's testing, the new DeepSeek flash model outperformed 5.6 Sol on Terminal-Bench 4.0 while being about 17x cheaper, according to the post.

## Evidence

Verbatim from https://www.baseten.co/blog/how-to-make-flash-models-even-faster-optimizing-deepseek-v41-flash-1017-fast:

> Many models are now orders of magnitude better than humans at kernel work. Profiling, rewriting, and benchmarking are well-defined and verifiable, making this the ideal task for an agent to hill-climb. We used our internal harness to optimize 80+ kernels and bottlenecks for DeepSeek V4.1 Flash 1017 to create the Fast variant. The model is on Artificial Analysis at 991 output tokens/s and the lowest time per task (Parento frontier). This post shares how we applied techniques like kernel fusion, Programmatic Dependent Launch (PDL), and host loop minimization.
> 
> Flash models now dominate the low-cost Pareto frontier.
> 
> These mid-sized models deliver fast, affordable intelligence suitable for most development work. For example, DeepSeek V4.1 Flash 1017 at native precision is a 552B-parameter MoE, with 8B active during prefill and 16B active during decode, but scores 32 on the new Artificial Analysis V5.0 Index, the same score as GLM-5.3. And it does so at approximately ⅓ the cost and 4x the speed of the larger model.

---

Record: https://forck.live/items/34078-how-to-make-flash-models-even-faster-optimizing-deepseek-v4-1-flash-1017
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md
