Recommended Models

Documentation Index

Fetch the complete documentation index at: https://docs.veri.studio/llms.txt

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

​

Overview

Veri supports two training methods across different model types:

• GRPO for language models (86 curated LLMs)
• Video Gen SFT for video generation models (10 curated models)

The right model depends on what you are optimizing for: iteration speed, output quality, structured behavior, or deployment constraints.

​

Language Models (GRPO)

For most teams, the best starting point is the smallest model that can already do the task reasonably well before fine-tuning.

​

Start small for fast iteration

If you are validating a workflow, reward function, or dataset shape, start with a smaller instruct model first. That gives you:

• cheaper experiments
• faster feedback loops
• easier debugging when outputs look wrong

​

Move up only when quality requires it

Use larger models when you need:

• stronger reasoning
• better function selection
• more reliable long-form outputs
• more headroom after fine-tuning

​

Video Generation Models (SFT)

Video gen SFT uses LoRA fine-tuning on supported diffusion models. Currently supported families:

​

Good Starting Questions

Ask yourself:

1. Do I need the cheapest iteration loop or the best possible quality?
2. Is the task language-based (GRPO) or video generation (SFT)?
3. Do I need a model that is easy to serve in my current stack after training?
4. How large a checkpoint can I realistically deploy?

​

Explore the Model Catalog

The model library is the best place to browse all supported models. Filter by LLM or Video Gen tabs. Open the model library

​

Next Steps

• Quickstart
• GRPO Training
• Supervised Fine-Tuning for video gen models