Overview

Cogito v1 Preview models were trained with iterated distillation and amplification, a self-improvement approach. Each can switch reasoning on with a system prompt.

Highlights

  • 3B to 70B sizes
  • Optional reasoning mode
  • Tool calling
  • Built on Llama 3.2/3.1 and Qwen2.5 bases

Typical uses

Local assistants that sometimes need deeper reasoning.

Things to consider

  • License depends on the base model of each size.
  • Download sizes refer to the default quantized versions in the Ollama library; full-precision weights are larger.

Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.