Skip to content
Model infoModelsSnapshot: 15 Oct 2026

ELYZA / ELYZA LLM

Llama-3-ELYZA-JP-120B

Research model enlarged from Llama 3 70B through depth upscaling and then trained for Japanese.

Language models (LLMs)
previewClosed weightsText generation
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
7 of 20 specification fields collected
Context windowNot collectedAwaiting source review
Maximum outputNot collectedAwaiting source review
Input billingNot collectedELYZA research
Output billingNot collectedNo inferred price

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum input tokens
Not collected
Maximum output
Not collected
Input modalities
TextSource for Input modalities (opens in a new tab)
Output modalities
TextSource for Output modalities (opens in a new tab)
Reasoning
Not collected
Tool calling
Not collected
Structured output
Not collected
Streaming
Not collected
Batch processing
Not collected
Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
122BSource for Parameters (opens in a new tab)

120B is the model name; slide 15 states 122 billion parameters.

Architecture
Dense transformer, depth upscaling from 80 to 140 layersSource for Architecture (opens in a new tab)
Languages
JapaneseSource for Languages (opens in a new tab)

Documented focus, not an exhaustive supported-language list.

Context caching
Not collected

No public prompt-caching contract, lifetime or cached-input rate verified. Speculative decoding is not a published cache discount.

Published price
Not collected

No numeric production price verified for this exact model. Historical free demos do not establish free production inference.

Base model
Meta-Llama-3-70B-InstructSource for Base model (opens in a new tab)
Published training data
Approximately 150B tokens of additional Japanese pretraining, followed by post-trainingSource for Published training data (opens in a new tab)

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

ELYZA research

Provider source (opens in a new tab)
API model ID
Not collected

Published model label is not a verified API request selector.

Access
Published research result. Exact API, downloadable derivative weights and a production launch are not verified.
Regions
Deployment-specific; no universal processing-region guarantee verified.

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Retention, training use, regional processing and enterprise terms have not been summarized for this endpoint.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

ELYZA research

Not verified
How it works
Not collected
Minimum reusable input
Not collected
Lifetime & refresh
Not collected
How to enable it
Not collected
Usage fields
Not collected
Billing rules
Not collected
Limits & data boundaries
Not collected

No verified cache rate for this endpoint. An empty rate is not a zero charge.

Caching reviewed 14 Oct 2026

What the measurements say

Japanese MT-Bench

ELYZA · vendor-reported
9.24score / 10

September 2024 provider presentation, slide 16. Earlier 70B dialogue/task scores are reused from its launch; not an independent evaluation or a universal model ranking.

Evaluation source (opens in a new tab)

ELYZA Tasks 100

ELYZA · vendor-reported
4.09score / 5

September 2024 provider presentation, slide 16. Earlier 70B dialogue/task scores are reused from its launch; not an independent evaluation or a universal model ranking.

Evaluation source (opens in a new tab)

Nejumi-v3 GLP average

ELYZA · vendor-reported
0.6898aggregate score

September 2024 provider presentation, slide 16. Earlier 70B dialogue/task scores are reused from its launch; not an independent evaluation or a universal model ranking.

Evaluation source (opens in a new tab)

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Provider identifies this separately trained derivative, but no matching public checkpoint was verified in ELYZA’s reviewed weight inventory. Closed-weight classification is inferred for this exact model, not the public upstream Llama or Qwen weights. Research access and redistribution terms remain unverified.

Classification evidence (opens in a new tab)

Lifecycle dates

Presented at GENIAC final report10 Sept 2024Source (opens in a new tab)

Sources reviewed