Comparative Review of LoRA Model Training
mesmerlord · reddit · 2026-07-14
The author spent 4 days testing LoRA training across various image models, comparing Krea, Ideogram, Flux 1 Dev, Flux 2 Dev, Flux 2 Klein, and Z Image.
The findings rank Ideogram's training performance as the best, followed by Flux 1 Dev, Z Image, Flux 2 Klein, and Krea, with Flux 2 Dev performing the worst. Additional insights include:
- Using Claude to automate the training pipeline and error fixing saved significant manual tuning effort.
- Training stability varied notably across different ethnicities and skin tones, with Black and South Asian samples being more prone to distorted results.
- Flux 2 Dev incurred high training costs and yielded subpar results due to encoder-related issues.
More from coding & agent
- Codex helps build Valdiluce, an open-world game with climbing, gliding and gondolas — Dimillian · 2026-07-22
- HeyGen adds a media-sourcing skill for coding agents with 75k images and 10k tracks — HeyGen · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- LangSmith adds tracing for Pipecat, LiveKit, OpenAI Realtime, and Gemini Live — LangChain · 2026-07-22
- An MCP server signs every AI agent tool call into a verifiable Merkle chain — Funky_Chicken_22 · 2026-07-22
- Annotated transcript of a Claude Code team interview is now available — trq212 · 2026-07-22