LlamaFactory Releases

This release introduces a comprehensive token usage tracking system with deduplication and database management, alongside improvements to metrics reingestion and CLI utilities. It also includes various daemon enhancements and dependency updates.

Aug 26, 2026

This release focuses on improving Windows and WSL installation workflows, normalizing git revisions, and isolating telemetry from core daemon processing. It also adds support for models.dev pricing integration.

Aug 23, 2026

Adds primary support for the Qwen3.5/Qwen3.6 and Gemma4 model families while upgrading the codebase to be compatible with Transformers v5. Includes support for multiple new models and training techniques.

May 30, 2026

Year-end release marking 2025 with major infrastructure changes including repository renaming, Python version updates, and migration to uv package manager. Adds support for new training methods like OFT, KTransformers backend, and Transformers v5.

Dec 31, 2025

Major release adding support for latest model families including Llama 4, Gemma 3, Qwen3, and InternVL3 with new multimodal capabilities, SGLang inference, and official Docker image.

Jun 16, 2025

Release focused on new optimizers, experiment tracking, and expanded model support including DeepSeek R1, Qwen2.5-VL, and various vision-language models with Ray Trainer integration.

Mar 11, 2025

Vision model focused release adding Llama-3.2-Vision, LLaVA-NeXT, Pixtral support along with Qwen2.5 Coder and gradient loss fix for transformers 4.46.

Nov 24, 2024

Major release introducing Qwen2-VL multimodal support, Liger-Kernel, Adam-mini optimizer, and expanded vision-language model training capabilities with RLHF/DPO approaches.

Sep 8, 2024

Introduces neat packing for contamination-free training, split evaluation capabilities, and support for HQQ/EETQ quantization methods. Also adds ZeRO-3 support for BAdam.

Jul 18, 2024

Adds support for GLM-4 tools and parallel function calling, introduces PiSSA fine-tuning, and supports DeepSeek-Coder-V2 models. Includes new datasets for supervised fine-tuning.

Jun 19, 2024

A patch release addressing compatibility issues with Unsloth+DoRA, PyTorch versions in Docker, and Windows installation. Fixes LongLoRA implementation issues.

Jun 10, 2024

Introduces GLM-4 and Qwen2 model support, adds KTO and SimPO algorithms, and significantly enhances the LlamaBoard Web UI with single-node distributed training capabilities.

Jun 7, 2024

A major refactor that shifts the architecture towards CLIs and YAML configurations, renames core files, and adds Ascend NPU and Yi-VL model support.

May 15, 2024

Adds LLaVA Multimodal LLM support, 2x faster generation via UnslothAI optimization, and native Transformers and vLLM inference support for LLaVA models.

Apr 27, 2024

Adds support for Meta Llama-3 models and significantly extends QLoRA context length capabilities to 56,000 tokens. Introduces BAdam and Mixture-of-Depths training algorithms.

Apr 21, 2024

This release introduces support for the ORPO algorithm and the Qwen1.5-32B model family, alongside improvements to quantization and dataset management.

Apr 11, 2024

A patch release fixing a critical bug in optimizer and scheduler creation when using DeepSpeed, alongside several other bug fixes.

Mar 29, 2024

A major release featuring a research paper, GaLore algorithm for memory efficiency, FSDP+QLoRA support, and vLLM integration.

Mar 25, 2024

Adds support for DoRA and QLoRA for AWQ/AQLM quantized models, alongside Gemma model support.

Feb 28, 2024

Introduces block expansion for LLaMA Pro, Qwen1.5 models, and fixes for DeepSpeed Zero3 with MoE models.

Feb 20, 2024