Skip to content
← Models
Open weightsKakao/10 trustMetrics-checked

Kanana-Nano-1.5-3B

Kakao's upcycled 3B nano model, the seed for its first MoE release.

Visit official page →

About

Kanana-Nano-1.5-3B is Kakao's 3B-scale open-source language model, developed by upcycling to save training time and cost. Despite only 3B parameters, Kakao reports its performance is equivalent to or exceeds Kanana-1.5-8B.

It served as the seed for Kanana-1.5-15.7B-a3B, Korea's first open-source MoE, whose expert layers were created by replicating the nano model's MLP layers.

Like the rest of the Kanana 1.5 generation, it was released under Apache 2.0, permitting commercial deployment.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

July 1, 2025

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

language-modelingkorean-aion-devicecost-efficient-inference

Trust score — 7/10

/10 trustMetrics-checked

Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 116; Trending score 66/100 from researcher assessment.

Sources