Models & assistants Model or AI service
ELYZA adapts Llama 3 8B for Japanese
ELYZA’s 8B model starts from Meta Llama 3 8B Instruct and adds Japanese training. The developer published downloadable weights and documented the base model. [1][2]
- License
- Meta Llama 3 Community License for the cited checkpoint. [2]
What the model does
On June 26, 2024, ELYZA described its Japanese Llama 3 work and made the 8B model public. The company reported results from a separate 70B variant as part of the same announcement. [1][2]
Why it matters
A Japanese-adapted checkpoint lets developers compare an existing model with a version trained further on regional language data. The shared lineage makes that comparison easier to interpret. [1][2]
Who can use it
Developers and researchers can evaluate the downloadable 8B checkpoint on Japanese text. The company also lists alternate weight formats for the 8B version. [1][2]
Key dates
- ELYZA announces the public 8B checkpoint [1]
Ways to use it
Compare how ELYZA and the original base respond to Japanese instructions or summarize the same source text. Look for preservation of meaning and names as well as fluent phrasing. [1][2]
Model limitations
The 8B checkpoint uses the Meta Llama 3 Community License. The developer reports results for more than one size, so check that a benchmark refers to the model under evaluation. [1][2]
Sources and verification
Catalog evidence checked How we verify records
- Primary source ELYZA’s June 2024 public 8B release and Japanese-language model report
- Primary source elyza/Llama-3-ELYZA-JP-8B
- Primary source Hugging Face registry metadata for elyza/Llama-3-ELYZA-JP-8B
Related profiles
Models & assistants
GPT-6.2 Astra and Sol
OpenAI
Astra targets demanding projects. Sol targets everyday tasks. Both support text and image input in the API.
Models & assistants
MAI Code 2 and Code 2 Flash
Microsoft AI
Code 2 targets complex software work. Flash targets frequent edits, explanations and routine fixes.
Models & assistants
MAI Realtime 1
Microsoft AI
Users can try a MAI Labs demo. Developers can register for a limited Foundry preview.