Vendor · Language models · Audio models · Specialist models
NVIDIA has 29 models in the directory: 26 language models, 2 audio models and 1 specialist model.
NVIDIA supplies GPUs and CUDA software for AI development. OpenAI names NVIDIA as a strategic partner. Anthropic and NVIDIA announced an investment commitment of up to $10B in November 2025; this is an announced maximum, not evidence that the full amount has been paid.
In the directory: 26 language models, 2 audio models and 1 specialist model. 28 of them with open weights that can run locally.
Every model there ever was. Models with a profile lead to every benchmark, price and source.
| Nemotron 3.5 Bolt E4B | Oct 23, 2026 | Open weights | n/a | n/a | n/a | 66K | 53 |
| Nemotron 3.5 Bolt 74B A9B | Oct 15, 2026 | Open weights | n/a | n/a | n/a | 262K | 45 |
| Nemotron 3.5 Super VL 120B A12B | Oct 15, 2026 | Open weights | n/a | $0.15 / $0.50 | 438 | 1.05M | 52 |
| Nemotron 3.5 Lightning | Aug 11, 2026 | Active | below 1 | $0.06 / $0.16 | 298 | 1M | 49 |
| Nemotron 3.5 Lightning 30B A3B | Aug 11, 2026 | Active | n/a | n/a | n/a | 262K | 23 |
| Nemotron 3 Ultra | Jun 4, 2026 | Active | n/a | $0.60 / $2.40 | 163 | 1M | 53 |
| Nemotron 3 Ultra 550B A55B | Jun 4, 2026 | Active | n/a | n/a | 198 | 262K | 20 |
| Nemotron 3 Nano Omni 30B A3B Reasoning | Apr 28, 2026 | Active | n/a | n/a | 229 | 256K | 45 |
| Nemotron Cascade 2 30B A3B | Mar 19, 2026 | Active | n/a | n/a | n/a | 262K | 38 |
| Nemotron VoiceChat | Mar 16, 2026 | Active | n/a | n/a | n/a | n/a | 31 |
| Nemotron 3 Super | Mar 11, 2026 | Active | n/a | $0.08 / $0.45 | 99 | 1M | 46 |
| Nemotron 3 Super 120B A12B | Mar 11, 2026 | Active | n/a | n/a | 162 | 262K | 18 |
| Nemotron 3 Nano | Dec 15, 2025 | Active | n/a | $0.05 / $0.20 | 223 | 1M | 44 |
| Nemotron 3 Nano 30B A3B | Dec 15, 2025 | Active | n/a | n/a | 274 | 262K | 18 |
| Nemotron Nano 12B v2 VL | Oct 28, 2025 | Active | n/a | n/a | 204 | 131K | 24 |
| Nemotron Nano 2 VL 12B | Oct 28, 2025 | Open source | n/a | $0.20 / $0.60 | 72 | n/a | 33 |
| Nemotron Nano 2 9B | Aug 18, 2025 | Open source | n/a | $0.06 / $0.23 | 143 | 128K | 45 |
| Nemotron Nano 9B v2 | Aug 18, 2025 | Active | n/a | n/a | 164 | 131K | 15 |
| Llama 3.3 Nemotron Super 49B v1.5 | Jul 25, 2025 | Active | n/a | n/a | n/a | 131K | 39 |
| Mistral Nemotron | Jun 11, 2025 | Active | n/a | n/a | n/a | 128K | 3 |
| Llama 3.1 Nemotron 70B Instruct | Apr 15, 2025 | Active | n/a | n/a | n/a | 128K | 35 |
| Llama 3.1 Nemotron Ultra 253B | Apr 7, 2025 | Active | n/a | n/a | n/a | 131K | 23 |
| Llama 3.3 Nemotron Super 49B v1 | Mar 18, 2025 | Active | n/a | n/a | n/a | 131K | 34 |
| Nemotron Mini 4B Instruct | Aug 21, 2024 | Active | n/a | n/a | n/a | 4K | 35 |
| Nemotron-4 340Bno profile | June 2024 | No status data | n/a | n/a | n/a | n/a | n/a |
| Megatron-Turing NLGno profile | January 2022 | No status data | n/a | n/a | n/a | n/a | n/a |
| NVIDIA Nemotron 3.5 Transcribe | Oct 28, 2026 | Open weights | n/a | 48 |
| Nemotron-3-Diarization | Sep 23, 2026 | Open weights | n/a | 57 |
| Kumo-Tabular | Sep 29, 2026 | Open weights | n/a | 53 |
Audio model · Oct 28, 2026
Nemotron 3.5 Transcribe is an open-weight speech model that combines audio recordings with text instructions. A parallel audio and speaker encoder feeds a Mamba-2 and Transformer mixture-of-experts decoder with 34 billion total parameters and about 4.3 billion active parameters. NVIDIA lists 40 transcription languages and recordings up to one hour long.
Language model · Oct 23, 2026
Nemotron 3.5 Bolt E4B has 8 billion total parameters and 4 billion effective parameters. It builds on Gemma-4-E4B-it. The BF16 config specifies 65,536 context tokens, and OpenMDW 1.1 governs the model weights.
Language model · Oct 15, 2026
Nemotron 3.5 Bolt 74B A9B is a hybrid MoE model with 74 billion total parameters and 9 billion active parameters. Its BF16 config specifies 262,144 context tokens. It accepts text, images, and video, returns text, and uses OpenMDW 1.1.
Language model · Oct 15, 2026
Nemotron 3.5 Super VL is a multimodal MoE model with 120 billion total parameters and 12 billion active parameters. NVIDIA combines Mamba-2, Transformer layers, 512 routed experts, and a vision encoder. It accepts text, images, and video and returns text.
Specialist model · Sep 29, 2026
Kumo-Tabular is NVIDIA’s tabular foundation model for classification and regression. You provide labeled context examples and query rows, and the model returns class probabilities or regression values. It works on structured tables and is not a System One chat model.
Audio model · Sep 23, 2026
Nemotron-3-Diarization assigns audio segments to up to eight speakers. NVIDIA’s card describes streaming and offline modes, 10 millisecond output resolution, and the OpenMDW-1.1 license.
Language model · Aug 11, 2026
Nemotron 3.5 Lightning uses a 30-billion-parameter MoE with 3 billion active parameters and multi-token prediction. NVIDIA emphasizes English and several programming languages. The checkpoint is therefore aimed at coding and controllable reasoning rather than chat alone.
Language model · Aug 11, 2026
Nemotron 3.5 Lightning 30B A3B is the BF16 checkpoint in the Lightning family. It activates 3 billion of 30 billion parameters and uses the hybrid Mamba2, MoE, and attention architecture with multi-token prediction. The model returns text.
Language model · Jun 4, 2026
Nemotron 3 Ultra combines 550 billion total parameters with 55 billion active parameters in NVIDIA's LatentMoE architecture. Mamba2, MoE, attention, and multi-token prediction form the technical basis. The checkpoint supports ten languages and returns text only.
Language model · Jun 4, 2026
Nemotron 3 Ultra 550B A55B is NVIDIA's BF16 checkpoint in the Ultra tier. Its hybrid LatentMoE architecture combines Mamba2, MoE, and attention with multi-token prediction. It has 550 billion total parameters and activates 55 billion.
Language model · Apr 28, 2026
Nemotron 3 Nano Omni 30B A3B Reasoning adds native audio input to NVIDIA's Nano line. The model handles text, images, video, and audio in one reasoning checkpoint. It activates 3 billion of 30 billion parameters.
Language model · Mar 19, 2026
Nemotron Cascade 2 30B A3B is a post-trained reasoning checkpoint. NVIDIA describes it as developed from a base model in the Nemotron 3 Nano line and provides separate Think and Instruct modes. Its 30 billion total parameters and 3 billion active parameters combine sparse serving with deliberate control over planning.
15 more current models are in the model list above.
| Vendor | Models | Model types | Best language model | Top 10 | Open weights | New in 2026 |
|---|---|---|---|---|---|---|
| Anthropic | 33 | 33 language models | Claude Opus 5.5 (Rank 1 to 3) | 4 | 0 | 13 |
| OpenAI | 63 | 46 language models, 4 image models, 2 video models, 9 audio models, 2 specialist models | GPT-6 Astra (Rank 4 to 8) | 2 | 4 | 27 |
| 66 | 43 language models, 4 image models, 6 video models, 11 audio models, 2 specialist models | Gemini 4 Carbon (Rank 1 to 5) | 2 | 19 | 26 | |
| Meta | 32 | 29 language models, 1 image model, 1 video model, 1 specialist model | Muse Wonder Full (Rank 6 to 9) | 2 | 26 | 7 |
| DeepSeek | 23 | 23 language models | DeepSeek V4.1 Flash 1017 (Rank 13 to 20) | 0 | 23 | 9 |
| Moonshot AI | 12 | 12 language models | Kimi K3.1 Pro (Rank 10 to 13) | 0 | 10 | 7 |
| xAI | 16 | 12 language models, 2 image models, 2 video models | Grok 4.7 (Rank 18 to 24) | 0 | 1 | 12 |
| NVIDIA | 29 | 26 language models, 2 audio models, 1 specialist model | not ranked yet | 0 | 28 | 15 |
By release date. Grey: retired models.
15 releases since January 2026, by release date.
Prices per 1M tokens for the current language models, according to NVIDIA or the provider linked in each model profile.
| Current model | Input | Output | Context |
|---|---|---|---|
| Nemotron 3.5 Lightning | $0.06 | $0.16 | 1M |
| Nemotron 3 Nano | $0.05 | $0.20 | 1M |
| Nemotron Nano 2 9B | $0.06 | $0.23 | 128K |
| Nemotron 3 Super | $0.08 | $0.45 | 1M |
| Nemotron 3.5 Super VL 120B A12B | $0.15 | $0.50 | 1.05M |
| Nemotron Nano 2 VL 12B | $0.20 | $0.60 | n/a |
| Nemotron 3 Ultra | $0.60 | $2.40 | 1M |
The directory lists 29 NVIDIA models: 26 language models, 2 audio models and 1 specialist model. It counts every model NVIDIA ever released, including retired ones; 27 of them have their own profile. 27 are currently available, the newest being NVIDIA Nemotron 3.5 Transcribe, Nemotron 3.5 Bolt E4B and Nemotron 3.5 Bolt 74B A9B.
NVIDIA Nemotron 3.5 Transcribe, released on Oct 28, 2026. The timeline below shows every release by date.
Through the API per 1M tokens (input / output): Nemotron 3 Nano $0.05 / $0.20. The prices of all 7 current language models are in the price table.
Yes, 28 of the 29 models have open weights. Each model profile lists how much memory it needs.
Nemotron 3.5 Super VL 120B A12B with 438 tokens per second according to Artificial Analysis. A typical model manages 98.
The largest is Nemotron 3.5 Super VL 120B A12B with 1.05 M tokens (≈ 1,570 A4 pages). Every model's value is in the model list.
Data as of Nov 1, 2026