Saved in:
Bibliographic Details
Main Authors: Dao, Alan, Vu, Dinh Bach
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.22760
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909689206276096
author Dao, Alan
Vu, Dinh Bach
author_facet Dao, Alan
Vu, Dinh Bach
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_22760
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Jan-nano Technical Report
Dao, Alan
Vu, Dinh Bach
Computation and Language
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.
title Jan-nano Technical Report
topic Computation and Language
url https://arxiv.org/abs/2506.22760