Fireworks Ultra-fast

Product

High-speed inference runtime for latency-sensitive agent workloads.

Launched: Nov 3, 2026
Models: —
Modality: —
Open source: —
API available: —
Releases: 1

Releases

Nov 3, 2026Fireworks Ultra-fast enters public preview for high-speed agent inferenceFireworks Ultra-fast · 1,500 tokens per second 2026-11-02/2026-11-09Product launch

Fireworks opened Ultra-fast in public preview on NVIDIA Blackwell GPUs, demonstrating up to 1,500 tokens per second on DeepSeek V4.1 Flash.

Fireworks AI Product launchFireworks AI Fireworks Ultra-fast