Join us for our inaugural conference, Forge 2026

Fireworks Blog

FireRouter with Opus

Introducing FireRouter with Opus

FireRouter with Opus routes routine turns to open models and keeps Claude Opus 5.5 for the turns that need it. 57% lower cost per session, cache-aware routing.

Filters

Reinforcement learning: Why alignment of numerics and MoE routing matter

Reinforcement learning: Why alignment of numerics and MoE routing matter

Multi-region deployment

Inside Fireworks Multi-region Deployments: One Deployment, Global Scale

ARCv3 and the case for cross-region RL

Company News9/23/2026

Every byte counts: ARCv3 and the case for cross-region RL

Graph showing 40% fewer tokens than Kimi K3

Model Releases9/23/2026

Introducing Ember-1

Specialized Intelligence Index

Company News9/22/2026

Introducing The Specialized Intelligence Index

The frontier isn't a model. It's a router headline image

The frontier isn’t a model. It’s a router.

DeepSeek-V4.1-Flash on Fireworks Headline Image

Model Releases9/14/2026

DeepSeek-V4.1-Flash on Fireworks: Astra-level DeepSWE at 1/15th the cost

Fireworks x Genspark

Partner Announcements9/10/2026

Gen-1 Slides: Opus 5-level decks at a fraction of the cost

Making the leap to specia

Developer Experience9/10/2026

Making the leap to specialized intelligence

By Sinan Ozdemir

Training API now generally available

Company News8/31/2026

Train past the frontier: Training API now generally available

Deepseek V4 Headline Image

DeepSeek V4 Pro: Tops SWE-Bench & Cuts Cost per Task by 3x vs. Fable 5

CyberGym Blog Image Header

DeepSeek V4 Pro is Redefining Security Agent Economics

Harvey x Fireworks

Partner Announcements8/26/2026

Post-training Kimi K3 with Harvey for long-horizon legal work

Header Image Do Open Models Perform Silent Reasoning Before They Respond

Can open models carry readable silent signals before they speak? Reproducing J-Lens Readouts on Kimi K3 & Qwen3.5-9B

Muse Glimmer on Fireworks

Muse Glimmer from Meta on Fireworks: Ideal for your Always-On Agents

Voyage AI Models on Fireworks

Model Releases8/5/2026

Your AI performance stack is Fireworks + Voyage AI

FullFT vs LoRA training

Three Tests to Run Before You Switch from LoRA to FullFT

fine-tune embeddings summary

Fine-Tune Your Own Embedding Model for the Price of a Coffee

Benchmarks Showing Opus 5 vs Kimi K3

Kimi K3 on Fireworks: Frontier Intelligence You Can Own

Showcase Serverless Training How it works

Make Kimi K3 Yours: LoRA Training on Fireworks

fireworks and trilogy

Trilogy’s Playbook for Open-Weight Cybersecurity with Kimi K3