Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.fireworks.ai/llms.txt

Use this file to discover all available pages before exploring further.

Give Fireworks your data and configuration. The platform handles scheduling, training, checkpointing, and model output. Training data uses the OpenAI-compatible chat completion format, so existing OpenAI SFT datasets work with no conversion required.

​
Methods

Supervised Fine Tuning - Text

Train text models with labeled examples of desired outputs

Supervised Fine Tuning - Vision

Train vision-language models with image and text pairs

Direct Preference Optimization

Align models with human preferences using pairwise comparisons

Reinforcement Fine Tuning

Train models using custom reward functions for complex reasoning tasks

​
Free Reinforcement Fine-Tuning

Reinforcement Fine-Tuning (RFT) is free for models under 16B parameters. When creating an RFT job in the UI, filter for free tuning models in the model selection area on the fine-tuning creation page. If kicking off jobs from the terminal, you can find the model ID from the Model Library. Note: SFT and DPO jobs are billed per training token for all model sizes—see the pricing page for details.
When creating a Reinforcement Fine-Tuning job in the UI, look for the “Free tuning” filter in the model selection area: Free tuning filter in the model selection area For SFT and DPO pricing, see the pricing page.

​
Supported base models

Fireworks supports fine-tuning for most major open source models, including DeepSeek, Qwen, Kimi, Gemma, GLM, and Llama families. The same set of base models is available for SFT, DPO, and RFT — once a base model is supported, every managed fine-tuning method works against it. The table below is generated from the live training shape registry. The “Max supported context length” is the largest max_supported_context_length across all training shapes registered for that base model — use it as the upper bound when you set a per-job context length on firectl sftj create, firectl dpoj create, or RFT job creation.
Base modelMax supported context length
gemma-4-26b-a4b-it256K (262,144 tokens)
gemma-4-31b-it256K (262,144 tokens)
glm-5p1200K (200,000 tokens)
kimi-k2p5256K (262,144 tokens)
kimi-k2p6256K (262,144 tokens)
llama-v3p3-70b-instruct128K (131,072 tokens)
minimax-m2p5192K (196,608 tokens)
nemotron-nano-3-30b-a3b256K (262,144 tokens)
qwen3-235b-a22b-instruct-2507128K (128,000 tokens)
qwen3-30b-a3b128K (131,072 tokens)
qwen3-30b-a3b-instruct-2507128K (128,000 tokens)
qwen3-32b128K (131,072 tokens)
qwen3-4b64K (65,536 tokens)
qwen3-8b256K (256,000 tokens)
qwen3-vl-8b-instruct256K (262,144 tokens)
qwen3p5-27b256K (262,144 tokens)
qwen3p5-35b-a3b256K (262,144 tokens)
qwen3p5-397b-a17b256K (262,144 tokens)
qwen3p5-9b256K (262,144 tokens)
qwen3p6-27b256K (262,144 tokens)
To browse the broader catalog (including non-tunable inference models), visit the Model Library for text models or vision models.

​
Tuning modes and context length

Managed fine-tuning supports both Low-Rank Adaptation (LoRA) and full-parameter tuning, depending on the model, method, and selected training shape. It also supports the full context lengths exposed by the available training shapes, matching the same long-context capabilities used by cookbook recipes. Choose LoRA when you want efficient adapter training and flexible deployment, including multiple LoRAs on a single base model deployment. Choose full-parameter tuning when you need to update all model weights for difficult reasoning, alignment, or domain adaptation tasks.
Deprecation notice: The deployedModel request key for routing to LoRA addons is deprecated and will not be supported for any new deployments. Please migrate to the model field with the <model_name>#<deployment_name> format described in Routing requests to LoRA addons.