Intelligence,
by Evolution.
Small models. An unconventional approach. We evolve models directly in the form you run, test the candidates, and keep the best.
- Qwen3 Max75.0%
- Claude Sonnet 474.3%
- Claude Opus 473.3%
- Vulcan 2.5B Coder70.0%
- gpt-oss-120b66.0%
- Gemini 2.5 Flash-Lite58.7%
- Mistral Medium 3.138.3%
- Mistral Large 338.0%
- GPT-4.134.7%
Built on a laptop.
Selected for what they can do.
Small by design.
Capable by selection.
Code, mathematics, and the space in between. Find your next model.
Vulcan 2.5B Coder
76.7% AIME 2026 · 70% AIME 2025A 2.5B coding and reasoning model. It won the selection league at 0.763 per question; its base scored 0.611.
Explore model01 From the leagueVulcan 2.5B Math
5.02 on the last four setsThe league's best score on the final question sets, which include mathematics. 0.728 per question overall.
Explore model02 Coming soonVulcan Large Coder
In the forgeA larger Vulcan for code, made with the same Integer Space Artificial Selection method.
Explore model03Coming soonVulcan Large Math
In the forgeA larger Vulcan for mathematics, made with the same Integer Space Artificial Selection method.
Explore model04Coming soonVulcan Embed
In the forgeAn embedding model from the Vulcan family, for search and retrieval.
Explore model05 Coming soonVulcan Embed Large
In the forgeA larger embedding model from the Vulcan family, for search and retrieval.
Explore model06