Fireworks AI unveils Forge products for company-specific AI models
Overview
Fireworks AI unveils new products at its Forge conference to help companies build AI models around their own data and tasks.
Fireworks Ultra-fast, running on NVIDIA Blackwell GPUs, is in public preview and reaches up to 1,500 tokens per second on DeepSeek V4.1 Flash, according to the company. Fireworks Sovereign is in private preview and brings Fireworks' inference engine to customers' cloud, neocloud, or data center environments.
The company also launches Nexus Desktop, which connects tools such as Codex app, Claude Desktop, and OpenCode to its model routing, with Pro plans at $8 per month and Teams at $10 per seat per month. Background Mode, for workloads that can wait, is offered at half price. Teams can get free training on Muse Glimmer 30B through November 30, covering supervised fine-tuning, direct preference optimization, and reinforcement learning.
Written by AI from the articles below · updated Nov 3, 4:31 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Fireworks AI BlogOfficialFireworks unveils Forge products for building company-specific AI models
AIFireworks launches Nexus Desktop, which connects tools such as Codex app, Claude Desktop, and OpenCode to its model routing, with Pro plans starting at $8 per month and Teams at $10 per seat per month. The company also unveils Fireworks Ultra-fast in public preview, reaching up to 1,500 tokens per second on DeepSeek V4.1 Flash, and Background Mode at half price for workloads that can wait. Fireworks also releases No Eval in private preview and Specialized Model Inference in public preview, starting with GLM 5.3 Flash.
Heat Trend
Not enough continuous observations to show a trend yet.