Home / Jobs / Fine-tune open models
Fine-tune open models
The best fine-tuning tool for training small open models on your data is Unsloth because it scores highest on our published formula (85.2/100): Open-source fine-tuning focused on speed and low memory. Very popular.
The job: “I want to fine-tune a small open model on my own data.” · Last checked
Unsloth
Open-source fine-tuning focused on speed and low memory. Very popular.
Not for: Large multi-node training runs.
Pricing: Open source. source, checked 2026-10-19
Visit Unsloth ↗Axolotl
Config-driven open-source fine-tuning toolkit.
Not for: Beginners who want notebooks over YAML configs.
Pricing: Open source. source
Score breakdown
| Tool | Fit (editorial) ×0.4 | Adoption (GitHub stars) ×0.25 | Momentum (release freshness) ×0.2 | Attention (HN, 30d) ×0.15 | Score |
|---|---|---|---|---|---|
| Unsloth | 5/5 → 1 | 78.2k★ → 0.917 | Oct 19, 2026 → 0.992 | 2 pts → 0.162 | 85.2 |
| Axolotl | 4/5 → 0.75 | 12.6k★ → 0.768 | Oct 16, 2026 → 0.963 | 3 pts → 0.204 | 71.5 |
Missing signals are dropped and the remaining weights re-normalized. See methodology. Signals fetched Oct 19, 2026.
FAQ
What is the best fine-tuning tool for training small open models on your data?
The best fine-tuning tool for training small open models on your data is Unsloth because it scores highest on our published formula (85.2/100): Open-source fine-tuning focused on speed and low memory. Very popular.
What is the best alternative to Unsloth?
Our runner-up is Axolotl (71.5/100). Config-driven open-source fine-tuning toolkit. Not for: Beginners who want notebooks over YAML configs.
Who should not use Unsloth?
Large multi-node training runs.
How much does Unsloth cost?
Open source. (checked 2026-10-19, source: https://github.com/unslothai/unsloth)
Related jobs
- Run LLMs locally: pick Ollama
- Self-host LLM serving at scale: pick vLLM
- One API for many models: pick OpenRouter
The weekly pick sheet
Once a week: what shipped, which picks changed, and why. Free.
Placeholder: signup isn't live yet and no email is collected. Follow RSS in the meantime.