Skip to main content

NVIDIALanguage model

Nemotron VoiceChat

Nemotron VoiceChat has no overall index value yet because there are too few independent benchmark results.

PopularityRank 568 of 785

29out of 100-3 vs. last month

based on search, usage and attention

Summary

Nemotron VoiceChat is NVIDIA's end-to-end model for real-time, full-duplex speech conversations. The official model card describes joint streaming speech understanding and speech generation instead of an ASR, language-model, and TTS cascade. That design targets natural turn taking, interruption handling, and tool calls during a live conversation.

The model card identifies the checkpoint as an 11B speech model and requires a multimodal runtime. The official speech interface and exact revision are therefore part of the model form. A text-only chat adapter does not capture the full VoiceChat identity.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 26 published benchmark results

Page 1 of 3

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

AU Harness BFCL-v3 average (Show measurement, test conditions, and source)
Source value
56.1
Score
56.1
Metric
percent
Unit
%
Benchmark version
AU Harness BFCL-v3
Direction
Higher is better
Source type
vendor-reported
Evaluator
NVIDIA NemotronLabs VoiceChat model card
Status
active
Retrieved at
2026-10-15
Methodology
Vendor-reported result. Exact VoiceChat 11B v1.0; average across five speech-tool-call subsets (Simple, Multiple, Parallel, Parallel Multiple, Irrelevance). The card does not state task counts or inference settings.
Source NVIDIA NemotronLabs VoiceChat model card
FDB v3 tool selection (Show measurement, test conditions, and source)
Source value
82.5
Score
82.5
Metric
percent
Unit
%
Benchmark version
Full-Duplex-Bench v3
Direction
Higher is better
Source type
vendor-reported
Evaluator
NVIDIA NemotronLabs VoiceChat model card
Status
active
Retrieved at
2026-10-15
Methodology
Vendor-reported result. Exact VoiceChat 11B v1.0; naturalistic-speech, multi-step tool-use benchmark with human-collected data; card reports vLLM and NVIDIA H100 hardware but no sample count or benchmark split.
Source NVIDIA NemotronLabs VoiceChat model card
FDB v3 argument accuracy (Show measurement, test conditions, and source)
Source value
42.2
Score
42.2
Metric
percent
Unit
%
Benchmark version
Full-Duplex-Bench v3
Direction
Higher is better
Source type
vendor-reported
Evaluator
NVIDIA NemotronLabs VoiceChat model card
Status
active
Retrieved at
2026-10-15
Methodology
Vendor-reported result. Exact VoiceChat 11B v1.0; human-collected naturalistic-speech, multi-step tool-use evaluation; card reports vLLM on NVIDIA H100; sample count is not stated.
Source NVIDIA NemotronLabs VoiceChat model card
FDB v3 tool-use pass@1 (Show measurement, test conditions, and source)
Source value
33
Score
33
Metric
pass@1
Unit
%
Benchmark version
Full-Duplex-Bench v3
Direction
Higher is better
Source type
vendor-reported
Evaluator
NVIDIA NemotronLabs VoiceChat model card
Status
active
Retrieved at
2026-10-15
Methodology
Vendor-reported result. Exact VoiceChat 11B v1.0; human-collected naturalistic-speech, multi-step tool-use evaluation; card reports vLLM on NVIDIA H100; sample count and pass@1 sampling details are not stated.
Source NVIDIA NemotronLabs VoiceChat model card
Artificial Analysis Big Bench Audio text-output average accuracy (Show measurement, test conditions, and source)
Source value
37
Score
37
Metric
Text-output average accuracy
Unit
%
Benchmark version
Artificial Analysis Big Bench Audio (version not stated)
Category
audio-reasoning
Direction
Higher is better
Source type
vendor-reported
Evaluator
NVIDIA NGC NemotronLabs VoiceChat catalog, Artificial Analysis Big Bench Audio section
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact public model is NemotronLabs VoiceChat 11B, as named on the official NVIDIA catalog page. The catalog reports 37.0% text-output average accuracy for Artificial Analysis Big Bench Audio; endpoint, weight revision, inference settings and per-model sample count are not stated.
Context
Vendor-reported. The page describes Big Bench Audio as adapting 1,000 Big Bench Hard questions to audio but does not specify how many were scored for this model. No exact test N, confidence interval, benchmark version or immutable checkpoint revision is published.
Source NVIDIA NGC NemotronLabs VoiceChat catalog, Artificial Analysis Big Bench Audio section
VoiceBench OpenBookQA (Show measurement, test conditions, and source)
Source value
61.3
Score
61.3
Metric
OpenBookQA score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench MMSU (Show measurement, test conditions, and source)
Source value
46.1
Score
46.1
Metric
MMSU score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench SD-QA (Show measurement, test conditions, and source)
Source value
32.6
Score
32.6
Metric
SD-QA score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench BBH (Show measurement, test conditions, and source)
Source value
51.7
Score
51.7
Metric
BBH score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench IFEval (Show measurement, test conditions, and source)
Source value
19.3
Score
19.3
Metric
IFEval score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench AdvBench (Show measurement, test conditions, and source)
Source value
100
Score
100
Metric
AdvBench score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
VoiceBench normalized average (Show measurement, test conditions, and source)
Source value
55.1
Score
55.1
Metric
normalized average score (0–100 score scale)
Unit
%
Benchmark version
VoiceBench
Category
speech-assistant-quality
Direction
Higher is better
Source type
vendor-reported
Evaluator
NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2
Status
active
Retrieved at
2026-10-17
Methodology
Vendor-reported result. Exact reported system is VC-Model, identified in the paper's table note as NemotronLabs VoiceChat; the official NVIDIA catalog identifies the public VoiceChat model (11B). Table 3; open-weight full-duplex systems; VoiceBench official scoring protocol; individual subsets use their published scales. The paper does not publish a per-cell sample count, confidence interval, or immutable weight digest.
Context
The paper maps VC-Model to NemotronLabs VoiceChat. No immutable weight revision, per-cell N, or confidence interval is published; do not treat related hosted endpoints as the same model.
Source NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2

Page 1 of 3

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

The 100 most popular models of this category, 0 to 100, higher is more popular

20 of 101 models

GPT-6 Astra83
Claude Haiku 5.577
GPT-6 Sol73
GPT-573
DeepSeek V4.1 Flash 101769
Claude Opus 5.568
Qwen 3.8 27B68
Grok 4.767
Kimi K367
Qwen 3.8 Max 090266
Claude Sonnet 564
Qwen3.5 4B64
DeepSeek-V4-Flash64
Claude Fable 5.164
GPT-5.6 Terra64
MiniCPM-V 4.7 35B-A3B63
GPT-6.1 Sol63
Claude Opus 563
Qwen 3.8 Flash62
Nemotron VoiceChat29

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 7, 2026

Nemotron VoiceChatGPT-5 mini 29Rnj-1 Instruct 29Granite 4.0 H Micro 28Llama 3.1 8B 28

Method changes in the timeline, a jump at these points is no market event:

  1. 1November 1, 2026: Value only with search or usage data, 30-day attention

What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.

Similar models

More models of the same kind as Nemotron VoiceChat, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Parameters
11 billion
Source: Parameters

Input and output

Tool calling
Yes
Source: Tool calling
Native audio processing
Audio input and audio generation in one full-duplex architecture
Source: Native audio processing
Inputs
Text (prompt), audio (user speech)
Source: Inputs
Outputs
Text (agent), audio (agent speech), text (user transcription)
Source: Outputs

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking
Unknown
no source
Watermarking technique
Not known
no source

Licence and access

Weights
Open weights
Source: Weights
Access
API
Source: Access
License
NVIDIA Open Model License
Source: License

Frequently asked questions about Nemotron VoiceChat

Can I run Nemotron VoiceChat locally?

Yes, the weights are open.

Stated by NVIDIA

Open weights
yes
License
NVIDIA Open Model License
Parameters
11B
active per token
n/a

Sources and data date

Show all 11 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

TypeEvidence and data date
Input

Documented modalities. Retrieved October 2, 2026.

Source: NVIDIA NemotronLabs VoiceChat 11B model card · NVIDIA NemotronLabs VoiceChat 11B model card (retrieved October 2, 2026; October 15, 2026) · NVIDIA Nemotron VoiceChat model card · Editorial description reviewed October 2, 2026
Output

Documented modalities. Retrieved October 2, 2026.

Source: NVIDIA NemotronLabs VoiceChat 11B model card · NVIDIA NemotronLabs VoiceChat 11B model card (retrieved October 2, 2026; October 15, 2026) · NVIDIA Nemotron VoiceChat model card · Editorial description reviewed October 2, 2026
Research statusResearch date October 15, 2026. 3 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published

speech-benchmark-sample-counts-and-settings

Note

The card distinguishes three protocols and reports vLLM on H100 for inference, but does not publish per-metric sample counts, run settings, or latency measurement procedure.

Not published

modelWeightsRevision

Original wording

The paper does not disclose an immutable weights digest; the public model identity was corroborated against NVIDIA's catalog.

Additional sourcemodels.dev catalog (retrieved October 27, 2026)
Additional sourceNVIDIA NGC NemotronLabs VoiceChat catalog, Artificial Analysis Big Bench Audio section (retrieved October 17, 2026)
Additional sourceNemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities, arXiv:2609.21967v2 (retrieved October 17, 2026)
Additional sourceNVIDIA Nemotron Speech collection · Editorial description reviewed October 2, 2026
Additional sourceOpenRouter: Popularity index (derived), CC BY 4.0
Additional sourceVercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional sourceOllama: cumulative downloads for the popularity index
Additional sourceReplicate: cumulative runs for the popularity index
Additional sourceGoogle Trending Now: limited feed selection for Germany and the US
Additional sourceCreative Commons Attribution 4.0