English

Jan-nano Technical Report

Computation and Language 2025-07-16 v2

Abstract

Most language models face a fundamental tradeoff where powerful capabilities require substantial computational resources. We shatter this constraint with Jan-nano, a 4B parameter language model that redefines efficiency through radical specialization: instead of trying to know everything, it masters the art of finding anything instantly. Fine-tuned from Qwen3-4B using our novel multi-stage Reinforcement Learning with Verifiable Rewards (RLVR) system that completely eliminates reliance on next token prediction training (SFT), Jan-nano achieves 83.2% on SimpleQA benchmark with MCP integration while running on consumer hardware. With 128K context length, Jan-nano proves that intelligence isn't about scale, it's about strategy.

Cite

@article{arxiv.2506.22760,
  title  = {Jan-nano Technical Report},
  author = {Alan Dao and Dinh Bach Vu},
  journal= {arXiv preprint arXiv:2506.22760},
  year   = {2025}
}
R2 v1 2026-07-01T03:37:35.842Z