nvidia
NVIDIA: Nemotron 3.5 Lightning
nvidia/nemotron-3.5-lightning
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
nvidia/nemotron-3.5-lightning is NVIDIA's BF16 text-to-text mixture-of-experts model with 30B total and 3B active parameters, described by NVIDIA as reference weights for customization rather than the optimized inference release. The model card reports a hybrid Mamba-2, MoE and attention design, configurable reasoning, tool calling and structured-output support, multilingual coverage led by English, and up to 1M-token context on suitable deployments; OpenRouter lists a 262,144-token context and text input/output. Vendor-measured results cover reasoning, coding, agentic, instruction-following and long-context tasks.[1][2]
Strengths
- Vendor-reported hybrid architecture combines Mamba-2, expert routing and attention while activating only a fraction of total parameters, an efficiency-oriented design for specialized or high-throughput work.[1][2]
- This BF16 checkpoint is explicitly framed as a customization base for SFT, RL, distillation, domain adaptation and user-produced quantized variants.[2]
- Catalog parameters include tools, structured outputs, sampling controls and reasoning options; NVIDIA also documents an on/off reasoning mode.[1][2]
Limitations
- The evidence pack contains a marketplace catalog and NVIDIA's own card, but no independent model card or third-party evaluation; published benchmark numbers are vendor-measured, so validate quality, latency and cost on the target workload.[1][2]
- This BF16 variant is not NVIDIA's recommended route for optimized production inference; the card points to a separate NVFP4 release, and its properties should not be assumed here.[2]
- Context depends on deployment: a single 80GB H100 is validated at 256K in BF16, while the largest documented configurations reach 1M; OpenRouter exposes 262,144.[1][2]
Where to use it
- A reasonable candidate for teams planning post-training, domain adaptation, distillation or custom quantization from full-precision weights.[2]
- Worth evaluating for text-only agents, chat, RAG and instruction-following prototypes that need tool calls or structured output; production throughput should be tested before commitment.[1][2]
- Potentially useful for long-context research, retrieval or multi-document aggregation experiments when the selected hardware can support the required context window.[1][2]
Sources
- [1] OpenRouter catalog · NVIDIA: Nemotron 3.5 LightningSources checked: Nov 13, 2026
- [2] Model documentation · NVIDIA: Nemotron 3.5 LightningSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
NVIDIA Nemotron 3.5 Lightning is an open mixture-of-experts model from NVIDIA, with 3B active parameters out of 30B total. It is suited for high-throughput agentic workloads and specialized tasks that...
Specifications
- Model ID
- nvidia/nemotron-3.5-lightning
- Canonical ID
- nvidia/nemotron-3.5-lightning-20260807
- Creator
- nvidia
- Added to OpenRouter
- 11.08.2026
- Context length
- 262,144
- Modality
- text->text
- Input modalities
- text
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.070 / out $0.200
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | CC — below the B band | reasoning: optional · parameters: include_reasoning, reasoning |
| Coding | CC — below the B band | — |
| Agents | CC — below the B band | parameters: structured_outputs, tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · parameters: include_reasoning, reasoning |
| Document analysis | No measurement | — |
| Vision | No measurement | — |
| Image generation | No measurement | — |
| Audio | No measurement | — |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Intelligence index | 5.5 | CC — below the B band | 13.11.2026 |
| Artificial Analysis | Coding index | 26.8 | CC — below the B band | 13.11.2026 |
| Artificial Analysis | Agentic index | 3.5 | CC — below the B band | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| Darkbloomint4 | 262,144 | $0.025 | $0.120 | — | — | 100% |
| Fireworks | 262,144 | $0.050 | $0.200 | — | — | 100% |
| DeepInfrabf16 | 262,144 | $0.060 | $0.160 | — | — | 100% |
| CoreWeavebf16 | 262,144 | $0.070 | $0.200 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.