A 3‑billion‑parameter Japanese‑focused LLM, Sarashina2.2‑3B, trained in three phases—including synthetic math/coding data and fine‑tuning on small application data—outperforms larger models on Japanese QA, math and coding benchmarks, but lacks instruction tuning and can produce biased or nonsensical outputs.
Description in other languages
Statistics
Statistics as of Nov 16, 2026
Hugging Face downloads (last 30 days)
2,000
Hugging Face likes
22
Listed
Sep 14, 2026
Downloads (30-day window) over time
Show as a table
| Month | DL (30d) |
|---|---|
| Dec 2025 | — |
| Jan 2026 | — |
| Feb 2026 | — |
| Mar 2026 | — |
| Apr 2026 | — |
| May 2026 | — |
| Jun 2026 | — |
| Jul 2026 | — |
| Aug 2026 | — |
| Sep 2026 | 1,000 |
| Oct 2026 | 1,000 |
| Nov 2026 | 2,000 |
Downloads over the 30 days before the last snapshot of each month.
Popular in the same category
- deberta-v3-base-japaneseku-nlp13kJapanese DeBERTa V3 base pre‑trained on 540 B tokens from LLM‑jp v1.0, trained with a modified DeBERTa V3 setup, uses a unigram byte‑fallback tokenizer (no morphological analyzer), and is fine‑tuned for JGLUE NLU tasks.
- Llama-3.1-Swallow-8B-Instruct-v0.5-ggufmmnga7kGGUF conversion of Llama‑3.1‑Swallow‑8B‑Instruct‑v0.5 by tokyotech‑llm, incorporating TFMC/imatrix‑dataset‑for‑japanese‑LLM, with Build/Run instructions for llama.cpp.
- sarashina2.2-3b-instruct-v0.1-ggufmmnga4kA GGUF‑formatted conversion of sbintuitions’ sarashina2.2‑3b‑instruct‑v0.1 model built with TFMC/imatrix‑dataset‑for‑japanese‑llm, ready for CUDA‑enabled inference via llama.cpp.
- llm-jp-3.1-13b-instruct4-ggufmmnga4kA GGUF‑formatted version of llm‑jp‑3.1‑13b‑instruct4, built with imatrix data from TFMC/imatrix‑dataset‑for‑japanese‑llm, usable via llama.cpp on CUDA for Japanese language tasks.
- Llama-3-ELYZA-JP-8B-ggufmmnga4kGGUF‑converted Llama‑3‑ELYZA‑JP‑8B by elyza, built with the TFMC/imatrix‑dataset‑for‑japanese‑LLM, ready for use with llama.cpp.
- llm-jp-3.1-1.8b-instruct4-ggufmmnga3kA GGUF‑formatted version of llm‑jp‑3.1‑1.8b‑instruct4, built from the TFMC/imatrix‑dataset‑for‑japanese‑llm and usable with llama.cpp’s CUDA build for Japanese LLM inference.
Resources in similar research fields
- auto-wiki-qacl-nagoya747AutoWikiQA is the largest free Japanese QA dataset, providing over 2.3 million handcrafted‑filtered question–answer pairs automatically generated from Wikipedia with Swallow‑MX and LLMs (without rule‑based templates) to support knowledge‑teaching and retrieval‑augmented generation applications.HF datasetShared: Large language models・QA & reasoning
- nri-fin-reasoningnri-ai707Japanese instruction dataset with 632,636 multi‑turn samples (~6.35 B tokens) and GPT‑OSS‑120b reasoning traces for open‑ended, math, writing, and MCQA tasks across 135 finance topics plus 20 general ones, designed to fine‑tune LLM reasoning in finance.HF datasetShared: Large language models・QA & reasoning
- japanese-gpt-neox-smallrinna656kA 12‑layer, 768‑hidden Japanese GPT‑NeoX model trained on CC‑100, C4, and Wikipedia, compatible with Huggingface, with an optional toy prefix‑tuning weight that forces each sentence to end with a smiling‑face emoji.Shared: Large language models
- Spark-X2.5-4B-Heretic-jp-ggufsoyaakinohara589kJapanese‑adapted Spark‑X2.5‑4B model (Heretic JP) with refusal attenuation, pretraining, instruction tuning and tool‑calling repair, available in GGUF quantizations from BF16 to Q4_K_M (~2.5 GiB), usable via llama.cpp for 1,048,576‑token context.Shared: Large language models
- llm-jp-4-33b-thinking-ggufllm-jp271kLLM‑jp‑4‑33b‑thinking‑gguf is a 33 B Japanese language model from the National Institute of Informatics, pre‑trained with SFT and DPO (no RL), available for llama.cpp via an LLM‑jp fork, and includes detailed usage instructions in its cookbook.Shared: Large language models
- llm-jp-4.1-8b-thinkingllm-jp80kLLM‑jp‑4.1‑8b‑thinking is a 33 B transformer language model from NII’s LLM R&D Center, pre‑trained with mid‑training only and aligned via SFT/DPO (no RL), designed for Japanese NLP tasks and documented in a cookbook and technical blog.Shared: Large language models