# fal — How We Achieved ~1000 tok/s and 16x Throughput with DSpark for Ideogram V4 Prompt Expander

- Company: fal (fal.ai)
- Announced: 2026-07-08T15:14:35+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://blog.fal.ai/how-we-achieved-1000-tok-s-and-16x-throughput-with-dspark-for-ideogram-v4-prompt-expander/
- Record: https://forck.live/items/7995-how-we-achieved-1000-tok-s-and-16x-throughput-with-dspark-for-ideogram-v4
- Subject: fal platform
- Models affected: Qwen3.6, Qwen 3.5 397B MoE, Qwen 3.5 4B, 35B MoE variant, Ideogram V4

fal achieved 16 times higher throughput on Qwen3.6 for the Ideogram V4 prompt expander using DSpark on SGLang, reaching 700 tok/s after retraining a DFlash model on 250K prompt expansion samples from their fine-tuned 35B MoE model.

## Evidence

Verbatim from https://blog.fal.ai/how-we-achieved-1000-tok-s-and-16x-throughput-with-dspark-for-ideogram-v4-prompt-expander/:

> At @fal , we've achieved 16 times higher throughput on Qwen3.6 for a use-case that required high interactivity per user using DSpark on SGLang.

---

Record: https://forck.live/items/7995-how-we-achieved-1000-tok-s-and-16x-throughput-with-dspark-for-ideogram-v4
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md