README.md

metadata

title: README
emoji: 🦀
colorFrom: pink
colorTo: gray
sdk: static
pinned: false

We work on making small language models punch far above their weight while remaining lightweight and deployable on edge!

These are our models:

Astral 1.5

Astral is our open-data family, trained for STEM usage.

• Astral 1.5 4B (link)

Astral

Astral is our first open-data family, trained for STEM usage.

• Astral 14B (link)
• Astral 4B (link)
• Astral 0.6B Flash (link)
• Astral 4B Preview (link)

Astral-Coder

Astral models fine tuned for pure coding and agentic usage

• Astral 1B (link)
• Astral 0.6B Flash (link)

Synth

• Synth 2 (link)
• Synth 1.5 (link)
• Synth 1 (link)

RedTeamer

RedTeamer is a model that generates jailbreaks that can be used for testing against LLMs.

• RedTeamer 1 (link)

Koishi

Koishi is a data synthesis LLM that generates CoT traces for non-CoT examples.

• Koishi 1 (link)

Kiwi (Depreciated)

Kiwi is a family of models trained for general STEM tasks.

• Kiwi 1.7B Nano (link)
• Kiwi 4B Mini (link)
• Kiwi 8B (link)

Lumi (Depreciated)

Depreciated STEM tuned models based on Qwen2.5 3B

• Lumi 1.5 (3B) (link)
• Lumi 1 (3B) (link)