Fireworks opened a public preview for serving specialized LoRA adapters through its Serverless API, starting with GLM 5.3 Flash. Teams could invoke an adapted model with per-token pricing and without sizing a dedicated endpoint for every adapter.
Company / HubLLMCross-Modal
Fireworks AI
Also known as: Fireworks
Plain English
Fireworks AI hosts models behind APIs and also lets teams run their own model deployments. A team can start on shared capacity or use a deployment configured for its workload.
Technical Overview
The platform offers shared serverless inference and configurable on-demand deployments. Deployment shapes provide presets for different latency, throughput, or cost priorities. The platform can also serve fine-tuned variants, including LoRA adapters on compatible base models.
Chronological Entries (2)
#2Release
AI-Assisted
#1Release
AI-Assisted
Fireworks introduced presets for on-demand deployments. Teams could choose a configuration tuned toward latency, throughput, or lower cost without manually selecting each serving parameter.
Direct Relations (2)
- Part ofInference Provider
- Alternative toBaseten