Fireworks Forge keynote: Sovereign stack, Ember next-gen targeting 30x efficiency, No-GPU
AccBalanced · x · 2026-11-04
Fireworks AI announced a wave of products at its Forge event: Fireworks Sovereign (the full Fireworks stack deployable in any environment); Ember next-gen — Ember-1, built on Kimi K3, delivered Kimi K3 quality with 40% fewer tokens by trimming unneeded reasoning, and the next version targets 30x more efficiency on US open models; No-GPU (pay-per-token training/inference); No-evals (evals built automatically from traces and product signals); Ultra-loop (recursive self-improvement for inference with a shared knowledge base); and Ultra-fast inference at 1,500–2,000 tok/s.
Related event: Fireworks unveils ultra-fast inference and sovereign AI stack(2 posts)→
More from Infra
- Sipeed's 10cm MaixStation Dock claims DGX Spark-class local AI inference, runs off a power bank — vincent_koc · 2026-11-04
- vLLM, Red Hat and NVIDIA to host open-source inference meetup in Madrid on Nov 12 — vllm_project · 2026-11-04
- vLLM v0.33.0 ships with 927 commits, 1.5–2x faster decode on GB300 — vllm_project · 2026-11-04
- Fireworks launches Ultra-fast inference up to 1,500 tok/s and half-price background mode at Forge — nicolechirps · 2026-11-04
- Fireworks event slides show sliding-window bit-sharing quantization trick — sophiamyang · 2026-11-04
- ThunderAgent merged into LLM-D: 2x agent serving throughput, TTFT from ~50s to 2-4s — simran_s_arora · 2026-11-04