Data as of October 29, 2026 · How the score is built
Nemotron 3.5 Transcribe 30B A4B
Decision readingNemotron 3.5 Transcribe 30B A4B is tracked, but not publicly ranked yet. Published speech or voice results from Nemotron 3.5 Transcribe launch evaluation appear below. They do not produce a weighted text-model score.
Released Oct 28, 2026 — see all recent releases
Voice benchmark evidence
Nemotron 3.5 Transcribe launch evaluation is an owner-defined voice evaluation. Its result stays separate from BenchLM's weighted text-model ranking.
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Base entry
Spec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- API model ID
- nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4BNVIDIA model documentation
- Context window
- 64KNVIDIA model documentation
- Maximum output
- Not sourced yet
- Knowledge cutoff
- Not sourced yet
- Input modalities
- Not sourced yet
- Output modalities
- Not sourced yet
- Parameters
- Not sourced yet
- Availability
- Released October 28, 2026 under the OpenMDW-1.1 license on Hugging Face, NGC, build.nvidia.com and NeMo Speech, in BF16 with FP8 and NVFP4 variants. It runs on one H100 80 GB with vLLM or Transformers, handles recordings up to one hour within a 65,536-token context, and supports multi-token-prediction speculative decoding plus an optional DFlash2 draft model. Reasoning is switchable in the chat template. NVIDIA describes it as offline: internal chunking does not make it a real-time streaming service.
- Cloud regions
- Not tracked yet
- Lifecycle
- Current
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Not documented in the pricing recordNVIDIA model documentation
- Self-host
- Open weights available; hardware estimate not sourced
- Rate limits
- Not tracked yet
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
We track Nemotron 3.5 Transcribe 30B A4B, but no weighted text-model benchmark result is published on the site yet. This page shows the metadata and separate protocol evidence we can verify now; a BenchLM score will appear only if compatible public evaluations land.
Nemotron 3.5 Transcribe 30B A4B is a open weight model with a 64K context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Released October 28, 2026 under the OpenMDW-1.1 license on Hugging Face, NGC, build.nvidia.com and NeMo Speech, in BF16 with FP8 and NVFP4 variants. It runs on one H100 80 GB with vLLM or Transformers, handles recordings up to one hour within a 65,536-token context, and supports multi-token-prediction speculative decoding plus an optional DFlash2 draft model. Reasoning is switchable in the chat template. NVIDIA describes it as offline: internal chunking does not make it a real-time streaming service.
Nemotron 3.5 Transcribe 30B A4B is NVIDIA’s offline speech-augmented language model built on Nemotron 3.5 Lightning 30B-A3B: a hybrid Mamba-Transformer mixture-of-experts model with a Transformer ASR encoder and a Sortformer speaker encoder, about 34B parameters with 4.3B active. It transcribes 40 languages verbatim or in written form, with context terms, speaker tags and word timestamps, and also translates speech, identifies languages and answers questions about audio. Its provider-reported speech results stay separate from text-model rankings.
Speech or voice evidence appears above; text-model benchmark coverage remains empty.
Last updated October 29, 2026. Runtime fields remain blank until a sourced snapshot exists.
Deployment options
Self-host and provider-specific paths stay separate from benchmark evidence so operating constraints are visible before a score becomes the whole decision.
Published weights are available, but BenchLM does not yet have a sourced parameter and VRAM profile for this exact model. Hardware cost estimates stay unavailable until that sizing record is complete.
Estimate VRAM from known parametersQuestions
How does Nemotron 3.5 Transcribe 30B A4B perform overall in AI benchmarks?
Nemotron 3.5 Transcribe 30B A4B has published speech or voice results in Nemotron 3.5 Transcribe launch evaluation. Those results remain separate from the weighted text-model ranking, so this profile does not assign a BenchLM score or overall rank. Use the linked evaluation's metric and run conditions to compare speech systems.
Is Nemotron 3.5 Transcribe 30B A4B open source?
Nemotron 3.5 Transcribe 30B A4B is an open-weight model from NVIDIA. Its weights can be downloaded for local or hosted deployment, subject to the published license. Open weight does not automatically mean open source: training data and training code may remain private, and commercial restrictions can still apply.
Does Nemotron 3.5 Transcribe 30B A4B have full benchmark coverage on BenchLM?
No. Nemotron 3.5 Transcribe 30B A4B's speech or voice results appear in Nemotron 3.5 Transcribe launch evaluation, but it has no published text-model benchmark rows. The voice evaluation does not fill the text-model coverage slots or produce an overall rank. Missing text coverage is not treated as a zero score.
What is the context window size of Nemotron 3.5 Transcribe 30B A4B?
Nemotron 3.5 Transcribe 30B A4B has a documented context window of 64K. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.