Fireworks Says Ultra-Fast Inference Hits 1,500 Tokens per Second
Summary
Fireworks launched Ultra-fast inference in public preview at its Forge conference, claiming up to 1,500 tokens per second on NVIDIA Blackwell GPUs. Nexus Desktop joins the Nexus CLI, with Pro plans starting at $8 per month. Fireworks also previewed task-specific evaluations and Fireworks Sovereign, which runs its inference engine inside customer infrastructure.
Key Points
- Fireworks claims a benchmarked 30-turn agent task drops from 1 minute to 8 seconds on Ultra-fast.
- Fireworks demonstrates the public preview on DeepSeek V4.1 Flash running on NVIDIA Blackwell GPUs.
- Fireworks targets agents that make multiple model calls, where each call adds latency for waiting users.