Fireworks on AI Knowledge Base

Account quotas

Mon, 01 Jan 0001 00:00:00 +0000

Account-wide request limits, spending tiers, budget controls, and on-demand GPU quotas

Agent Frameworks

Mon, 01 Jan 0001 00:00:00 +0000

Build production-ready AI agents with Fireworks and leading open-source frameworks

Agent Tracing

Mon, 01 Jan 0001 00:00:00 +0000

Understand where your agent runs and how tracing enables reinforcement fine-tuning

Anthropic compatibility

Mon, 01 Jan 0001 00:00:00 +0000

Use Anthropic SDKs with Fireworks, and understand the supported surface for the Anthropic-compatible Messages API.

Audit & Access Logs

Mon, 01 Jan 0001 00:00:00 +0000

Monitor and track account activities with audit logging for Enterprise accounts

Autoscaling

Mon, 01 Jan 0001 00:00:00 +0000

Configure how your deployment scales based on traffic

Basics

Mon, 01 Jan 0001 00:00:00 +0000

Understand the reinforcement learning fundamentals behind RFT

Batch API

Mon, 01 Jan 0001 00:00:00 +0000

Process large-scale async workloads

Build with Fireworks AI

Mon, 01 Jan 0001 00:00:00 +0000

Fast inference and fine-tuning for open source models

Checkpoints and Resume

Mon, 01 Jan 0001 00:00:00 +0000

Save training progress, resume from failures, and promote checkpoints to deployable models — driven by the recipe.

Claude Code

Mon, 01 Jan 0001 00:00:00 +0000

Use Claude Code with Fireworks AI models

Cleanup and Teardown

Mon, 01 Jan 0001 00:00:00 +0000

Delete trainer jobs and deployments after experiments to avoid leaked resources.

Client-side performance optimization

Mon, 01 Jan 0001 00:00:00 +0000

Optimize your client code for maximum performance with dedicated deployments

Completions API

Mon, 01 Jan 0001 00:00:00 +0000

Use the completions API for raw text generation with custom prompt templates

Concepts

Mon, 01 Jan 0001 00:00:00 +0000

This document outlines basic Fireworks AI concepts.

Cookbook Reference

Mon, 01 Jan 0001 00:00:00 +0000

Configuration classes, checkpoint utilities, and gradient accumulation normalization for cookbook recipes.

Cookbook: Distillation

Mon, 01 Jan 0001 00:00:00 +0000

Single-teacher OPD and routed multi-teacher policy distillation with cookbook recipes.

Cookbook: DPO

Mon, 01 Jan 0001 00:00:00 +0000

Direct Preference Optimization with pairwise data using the cookbook recipe.

Cookbook: Reinforcement Learning

Mon, 01 Jan 0001 00:00:00 +0000

Async RL on Fireworks — write a rollout function, the recipe owns the loop (gate, advantage, weight sync, KL/TIS, PPO, checkpoints). Runs async or fully synchronous.

Cookbook: SFT

Mon, 01 Jan 0001 00:00:00 +0000

Supervised fine-tuning via the cookbook’s sft_loop recipe.

Cookbooks

Mon, 01 Jan 0001 00:00:00 +0000

Interactive Jupyter notebooks demonstrating advanced use cases and best practices with Fireworks AI

Cost Estimator

Mon, 01 Jan 0001 00:00:00 +0000

Estimate and optimize the cost of your RFT training jobs

Courses

Mon, 01 Jan 0001 00:00:00 +0000

Standalone end-to-end examples showing how to use Fireworks to solve real-world use cases

Custom Models

Mon, 01 Jan 0001 00:00:00 +0000

Upload, verify, and deploy your own models from Hugging Face or elsewhere

Data Security

Mon, 01 Jan 0001 00:00:00 +0000

How we secure and handle your data for inference and training

Debug SFT tokenization

Mon, 01 Jan 0001 00:00:00 +0000

Download rendered token IDs and loss masks for supervised fine-tuning jobs.

Deploying Fine Tuned Models

Mon, 01 Jan 0001 00:00:00 +0000

Deploy one or multiple LoRA models fine tuned on Fireworks using live merge or multi-LoRA

DeploymentManager (Compatibility)

Mon, 01 Jan 0001 00:00:00 +0000

Legacy SDK reference for direct deployment lifecycle and weight-sync management.

Deployments

Mon, 01 Jan 0001 00:00:00 +0000

Configure and manage on-demand deployments on dedicated GPUs

Deployments Quickstart

Mon, 01 Jan 0001 00:00:00 +0000

Deploy models on dedicated GPUs in minutes

DeploymentSampler

Mon, 01 Jan 0001 00:00:00 +0000

Client-side tokenized sampling from inference deployments for training and evaluation.

Development Setup with Fireworks Docs MCP

Mon, 01 Jan 0001 00:00:00 +0000

Configure the Fireworks AI Docs MCP server for Claude Code and Cursor

Direct Preference Optimization

Mon, 01 Jan 0001 00:00:00 +0000

Embeddings & Reranking

Mon, 01 Jan 0001 00:00:00 +0000

Generate embeddings and rerank results for semantic search

Evaluators

Mon, 01 Jan 0001 00:00:00 +0000

Understand the fundamentals of evaluators and reward functions in reinforcement fine-tuning

Exporting Metrics

Mon, 01 Jan 0001 00:00:00 +0000

Export metrics from your dedicated deployments to your observability stack

Fire Pass Setup

Mon, 01 Jan 0001 00:00:00 +0000

Kimi K2.6 Turbo for personal agentic coding — Fire Pass (Early Access), $49 / month

FiretitanServiceClient & TrainingClient

Mon, 01 Jan 0001 00:00:00 +0000

Connect to a trainer endpoint and use the training client for forward/backward passes, optimizer steps, and checkpointing.

Fireworks Agent Overview

Mon, 01 Jan 0001 00:00:00 +0000

Describe what you want, approve the plan and cost, get a deployed fine-tuned model.

Fireworks Agent: Classification

Mon, 01 Jan 0001 00:00:00 +0000

Benchmark base models, fine-tune on labeled data, and pick the best classifier — automatically.

Fireworks Agent: Evaluator Authoring

Mon, 01 Jan 0001 00:00:00 +0000

Have Fireworks Agent generate a reusable evaluator from your dataset — for scoring candidates in an SFT sweep, or for use with Managed RFT.

Fireworks Agent: Preference Learning (DPO/ORPO)

Mon, 01 Jan 0001 00:00:00 +0000

Run preference fine-tuning end-to-end with optional base-model sweep, automatic pair generation, and pairwise evaluation.

Fireworks Agent: Supervised Fine-Tuning

Mon, 01 Jan 0001 00:00:00 +0000

Run end-to-end SFT with Fireworks Agent — dataset inspection, hyperparameter sweep, evaluator-guided selection, and a deployed winner.

FireworksClient

Mon, 01 Jan 0001 00:00:00 +0000

Account-level operations that don’t require a running trainer job.

GitHub Copilot

Mon, 01 Jan 0001 00:00:00 +0000

Use Fireworks AI models in GitHub Copilot Chat via a custom endpoint

Glossary

Mon, 01 Jan 0001 00:00:00 +0000

Definitions for key terms used across Fireworks AI documentation.

Incremental Snapshots (ARC2)

Mon, 01 Jan 0001 00:00:00 +0000

Build ARC2 incremental checkpoints, use per-file hints, and signal delta hot-loads for BYOT RL rollout integrations.

Inference Error Codes

Mon, 01 Jan 0001 00:00:00 +0000

Common error codes, their meanings, and resolutions for inference requests

Inference for RL Rollouts

Mon, 01 Jan 0001 00:00:00 +0000

Session affinity, weight-swap behavior, and MoE Router Replay for rollout traffic on Fireworks inference deployments.

Introduction

Mon, 01 Jan 0001 00:00:00 +0000

Fireworks Training API — custom training loops with full Python control over objectives, while Fireworks handles distributed GPU infrastructure.

Kimi K2 family

Mon, 01 Jan 0001 00:00:00 +0000

Using Kimi K2 family models in agentic and tool-calling workflows on Fireworks.

Ledger & Debugging for RL Rollouts

Mon, 01 Jan 0001 00:00:00 +0000

Inspect snapshot history, reset the ledger, and understand how in-flight requests behave during a weight swap.

Loss Functions

Mon, 01 Jan 0001 00:00:00 +0000

Built-in loss functions and custom objectives via forward_backward_custom.

Managed Fine-Tuning Overview

Mon, 01 Jan 0001 00:00:00 +0000

Fine-tune models with Fireworks-managed infrastructure — no custom code required.

Microsoft Foundry

Mon, 01 Jan 0001 00:00:00 +0000

Deploy frontier open models inside your Azure subscription, billed through Azure.

MLOps & Observability

Mon, 01 Jan 0001 00:00:00 +0000

Track and monitor your Fireworks AI deployments with leading MLOps and observability platforms

Monitor Training

Mon, 01 Jan 0001 00:00:00 +0000

Track RFT job progress and diagnose issues in real-time

OpenAI compatibility

Mon, 01 Jan 0001 00:00:00 +0000

openapi

Mon, 01 Jan 0001 00:00:00 +0000

Overview

Mon, 01 Jan 0001 00:00:00 +0000

Train models using reinforcement learning in minutes

Parameter Tuning

Mon, 01 Jan 0001 00:00:00 +0000

Learn how training parameters affect model behavior and outcomes

Performance benchmarking

Mon, 01 Jan 0001 00:00:00 +0000

Measure and optimize your deployment’s performance with load testing

Price comparison vs Tinker

Mon, 01 Jan 0001 00:00:00 +0000

Estimate the cost of multi-turn agentic RL rollouts on Fireworks compared to Tinker’s per-token pricing

Prompt caching

Mon, 01 Jan 0001 00:00:00 +0000

Python SDK

Mon, 01 Jan 0001 00:00:00 +0000

Quantization

Mon, 01 Jan 0001 00:00:00 +0000

Reduce model precision to improve performance and lower costs

Quickstart

Mon, 01 Jan 0001 00:00:00 +0000

Get a custom training loop running in minutes with the Fireworks Training API.

Reasoning

Mon, 01 Jan 0001 00:00:00 +0000

How to use reasoning with Fireworks models

Regions

Mon, 01 Jan 0001 00:00:00 +0000

Fireworks runs a global fleet of hardware on which you can deploy your models.

Reliability and Error Handling

Mon, 01 Jan 0001 00:00:00 +0000

Recommended patterns for timeouts, retries, and error handling when building production applications on the Fireworks API.

Remote Agent Quickstart

Mon, 01 Jan 0001 00:00:00 +0000

Train an SVG drawing agent running in a remote environment

Remote Environment Setup

Mon, 01 Jan 0001 00:00:00 +0000

Implement the /init endpoint to run evaluations in your infrastructure

Reserved capacity

Mon, 01 Jan 0001 00:00:00 +0000

Responses API

Mon, 01 Jan 0001 00:00:00 +0000

RFT parameters reference

Mon, 01 Jan 0001 00:00:00 +0000

Checkpoint, resume, and GRPO metrics fields for reinforcement fine-tuning recipes.

RL Rollouts with Your Own Trainer

Mon, 01 Jan 0001 00:00:00 +0000

Integrate an external RL trainer with Fireworks inference: hot-load new checkpoints from your bucket and run rollouts via the OpenAI-compatible API.

Routers

Mon, 01 Jan 0001 00:00:00 +0000

Distribute traffic across multiple deployments for A/B testing, traffic migration, and load distribution.

Saving and Loading

Mon, 01 Jan 0001 00:00:00 +0000

SDK-level reference for checkpoint save, load, weight sync, and promotion.

Secure Training (BYOB)

Mon, 01 Jan 0001 00:00:00 +0000

Fine-tune models while keeping sensitive data and components under your control

Serverless Overview

Mon, 01 Jan 0001 00:00:00 +0000

How Serverless inference works on Fireworks: serving paths, billing, request/response headers, prompt caching, model lifecycle, and when to choose Serverless over On-demand

Serverless Pricing

Mon, 01 Jan 0001 00:00:00 +0000

Per-token serverless pricing for text, vision, and embedding models, including Priority and Fast serving paths

Serverless Quickstart

Mon, 01 Jan 0001 00:00:00 +0000

Make your first Serverless API call in minutes

Serverless Rate Limits

Mon, 01 Jan 0001 00:00:00 +0000

Adaptive rate limits grow and shrink with your usage

Serverless Serving Paths

Mon, 01 Jan 0001 00:00:00 +0000

Standard, Priority, and Fast serving paths on Fireworks Serverless

Single-Turn Training Quickstart

Mon, 01 Jan 0001 00:00:00 +0000

Train a model to be an expert at answering GSM8K math questions

Speculative Decoding

Mon, 01 Jan 0001 00:00:00 +0000

Speed up generation with draft models and n-gram speculation

Structured Outputs

Mon, 01 Jan 0001 00:00:00 +0000

Enforce output formats using JSON schemas or custom grammars

Supervised Fine Tuning - Text

Mon, 01 Jan 0001 00:00:00 +0000

Supervised Fine Tuning - Vision

Mon, 01 Jan 0001 00:00:00 +0000

Learn how to fine-tune vision-language models on Fireworks AI with image and text datasets

Text Models

Mon, 01 Jan 0001 00:00:00 +0000

Query, track and manage inference for text models

The Cookbook

Mon, 01 Jan 0001 00:00:00 +0000

Ready-to-run training recipes for GRPO, DPO, SFT, and distillation built on top of the Training API.

Tool Calling

Mon, 01 Jan 0001 00:00:00 +0000

Connect models to external tools and APIs

TrainerJobManager (Compatibility)

Mon, 01 Jan 0001 00:00:00 +0000

Legacy SDK reference for service-mode trainer job lifecycle management.

Training and Sampling

Mon, 01 Jan 0001 00:00:00 +0000

End-to-end SDK walkthrough: bootstrap resources, train, checkpoint, and sample through a serving deployment.

Training Guide: UI

Mon, 01 Jan 0001 00:00:00 +0000

Launch RFT jobs using the Fireworks dashboard

Training Overview

Mon, 01 Jan 0001 00:00:00 +0000

Launch RFT jobs using the eval-protocol CLI

Training Overview

Mon, 01 Jan 0001 00:00:00 +0000

This overview frames Fireworks’ three fine-tuning paths — the autonomous Agent, semi-managed Managed Fine-Tuning, and the custom Training API — so it matters as the decision page before you commit compute. The key heuristic it offers is to reach for supervised fine-tuning when you have more than about a thousand quality labeled examples, and to switch to reinforcement fine-tuning for smaller datasets or reasoning-heavy tasks where ground-truth labels do not exist. A common mistake is defaulting to SFT on too little data. This is the Fireworks counterpart to Together AI’s fine-tuning flow; read the quickstart first if you are new to the platform.

Training Prerequisites & Validation

Mon, 01 Jan 0001 00:00:00 +0000

Requirements, validation checks, and common issues when launching RFT jobs

Training Shapes

Mon, 01 Jan 0001 00:00:00 +0000

Pre-configured GPU and model training profiles that simplify distributed training setup.

Understanding LoRA performance

Mon, 01 Jan 0001 00:00:00 +0000

Understand the performance impact of LoRA fine-tuning, optimization strategies, and deployment considerations.

Upload via REST API

Mon, 01 Jan 0001 00:00:00 +0000

Programmatically upload custom models using the Fireworks REST API

Use Fireworks Agent with Claude Code, Cursor, Codex, and other coding agents

Mon, 01 Jan 0001 00:00:00 +0000

Install the Fireworks Agent skill file once and drive end-to-end fine-tuning from your coding agent.

Using predicted outputs

Mon, 01 Jan 0001 00:00:00 +0000

Use Predicted Outputs to boost output generation speeds for editing / rewriting use cases

Using Secrets

Mon, 01 Jan 0001 00:00:00 +0000

Learn how to create secrets that can be utilized within your reward function.

Video & Audio Inputs

Mon, 01 Jan 0001 00:00:00 +0000

Query multimodal models to process video and audio content directly

Vision Inputs

Mon, 01 Jan 0001 00:00:00 +0000

Fine-tune vision-language models (VLMs) with the Training API using multimodal chat data containing images and text.

Vision Models

Mon, 01 Jan 0001 00:00:00 +0000

Query vision-language models to analyze images and visual content

Warm Start from Fine-Tuned Models

Mon, 01 Jan 0001 00:00:00 +0000

Continue training from a previously fine-tuned model with RFT

Weight sync

Mon, 01 Jan 0001 00:00:00 +0000

How a trainer’s updated weights reach the serving deployment during RL training.

Weighted Training

Mon, 01 Jan 0001 00:00:00 +0000

Control which samples have greater influence during RFT training

WeightSyncer (Legacy)

Mon, 01 Jan 0001 00:00:00 +0000

Backward-compatibility reference for the old standalone checkpoint-then-sync helper.

Which model should I use?

Mon, 01 Jan 0001 00:00:00 +0000

Find the best open models for your use case or migrate from closed source models like Claude, GPT, and Gemini

Zero Data Retention

Mon, 01 Jan 0001 00:00:00 +0000

Data retention policies at Fireworks