Nemotron-4-340B-Base

Which Nemotron-4-340B-Base does this page describe?

Nemotron-4-340B-Base is a model record from NVIDIA in the Nemotron family. Source

A separate version identifier is not documented in this record; the model name alone does not establish which snapshot a service runs.

How can you access Nemotron-4-340B-Base?

The catalog documents Open weights access for Nemotron-4-340B-Base. Source

As of the catalog check, this version was listed without confirmed general availability. Source

Open weights means the model weights are distributed for use under their stated terms. It does not establish a hosted API, an open-source license, or a hardware requirement.

What are Nemotron-4-340B-Base's context and output limits?

The documented context window is 4,096 tokens. Source

A separate maximum-output value is not established in this record. The context or input figure should not be used as an output allowance.

Context, input and output limits describe different constraints. Do not add the figures together to infer a larger window. Check the endpoint's accounting for conversation history, system instructions, tools, media and generated tokens; a documented maximum is not a guarantee of answer quality or a measured usable capacity.

What dates are documented for Nemotron-4-340B-Base?

Training-data cutoff: June 2023. Source This describes the reported data boundary, not the day training finished. The source labels this “Training-data cutoff as labeled in official model card.”

Retrieval or web tools can supply newer information at run time. That does not change the model's recorded training or knowledge dates, and this record does not establish that a particular endpoint includes those tools.

What license is recorded for Nemotron-4-340B-Base?

The recorded license label is nvidia-open-model-license. Source

Read the linked license and associated use terms for this exact version before redistribution, commercial deployment or fine-tuning. A label alone does not establish every permission, restriction or obligation, and it does not determine the terms of a separate hosted service.

What distinguishes Nemotron-4-340B-Base?

These details come from the linked publisher material for this model. Reported capabilities and training descriptions are publisher statements, not independent test results.

Checkpoint purpose

Pretrained 340-billion-parameter language model intended as a foundation for custom models. Source

Training period

Reported training period: between December 2023 and May 2024. This is not a data cutoff or public release date. Source

Architecture

Autoregressive Transformer decoder architecture. Source

Pretraining stages

The publisher reports eight trillion initial training tokens followed by one trillion continuous-pretraining tokens with a revised data mixture; the corpus includes over 50 natural and 40 coding languages. Source

BF16 deployment requirements

The published BF16 configurations use eight H200 GPUs, sixteen H100 GPUs, or sixteen A100 GPUs with 80 GB each. Source

Attention design

The card documents grouped-query attention and rotary position embeddings, with a 4,096-token training sequence. Training sequence length is kept distinct from service allowances. Source

What is documented about the Nemotron-4-340B-Base checkpoint?

This evidence refers to nvidia/Nemotron-4-340B-Base at recorded revision 5d954800b3e45c8234a9c40f4a1479240149eda8, checked October 23, 2026. Repository metadata

Listed library: nemo. Check that library's version and model support before loading the weights. Repository metadata

Repository access gate: false. This records the repository's gate setting when checked; licensing and use terms still apply. Repository metadata

Read the retrieved model card for setup and qualifications. Model card