Skip to content
← Models
Open weightsSB Intuitions/10 trustMetrics-checked

Sarashina2-Vision-8B

Japanese vision-language model pairing Sarashina2-7B with Qwen2-VL vision.

Visit official page →

About

Sarashina2-Vision-8B is a Japanese large vision-language model trained by SB Intuitions, the AI research arm of SoftBank. It combines the Sarashina2-7B language model with the image encoder of Qwen2-VL-7B, targeting Japanese document and image understanding.

The vendor reports top scores among Japanese VLMs on four benchmarks as of March 2025. The Sarashina family is trained with NVIDIA Nemotron tooling, and Sarashina3 mini was later selected by Japan's Digital Agency. Note the licensing: Sarashina models ship under the Sarashina Model NonCommercial License Agreement, so commercial use requires a separate arrangement.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

March 1, 2025

Benchmarks

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

Use cases

vision-languagejapanese-nlpdocument-understandingocr

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 128; Trending score 74/100 from researcher assessment.

Sources