Nemotron 3.5 Lightning
NVIDIA · released Aug 1, 2026 · nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16
- Type
- Open weightsCustom licence
- Params
- 31.6B
- Context
- 262K
active per word not recorded by us · about 197K words of context · download allowed, licence restricts use
Our take
Written Sep 15, 2026Nemotron 3.5 Lightning is a downloadable text model from NVIDIA with a 262,144-token request limit and broad Arena coverage. Its coding score is its standout measured skill, while creative writing lags well behind.
Pick this for long-context coding workflows where the Arena Coding score is your guide, or for budget-conscious inference across multiple providers. Use it when you need a very large request limit in a downloadable model. Skip it if you need a confirmed permissive licence, strong creative writing, or multimodal input.
The case for it
- Coding is its clear measured strength: 69.95 points above its overall Arena Text score.
- 262,144-token request limit, extremely large for a downloadable model.
- Seven hosted offers with output from well under a fifth of a cent per thousand tokens.
- A 272 tokens-per-second option exists for a modest input premium over the slowest host.
The case against it
- Creative writing is 78.59 points below its overall score and 148.55 points below its coding score.
- Licence terms are undisclosed despite weights being listed as downloadable.
- Throughput varies 90.7× between hosts at similar pricing, so provider choice matters enormously.
How good is it?
An open text model for everyday questions, though drafting and prose are weaker than most models here.
- drafts, rewrites and editingArena Creative Writing · 139th of 168
EverydayGeneral questions and everyday reasoning
Arena Text (overall)125th of 168 · 1347
CodingWriting and fixing code on its own
Arena Coding115th of 168 · 1418
Arena Coding is the only board that has scored it for this.
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent.
WritingDrafting and rewriting prose
Arena Creative Writing139th of 168 · 1277
Arena Creative Writing is the only board that has scored it for this.
These tests check whether a model follows instructions — a precondition for all the work above, but not a measure of how well that work is done.
Every published score for this model6 scoresEvery figure we hold, from 6 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
GeForce RTX 4090 · 24 GB
Borderline fit on an estimated size. It leaves 1 GB spare on a size we calculated rather than measured, and a 10% error either way would change the answer.
Comfortable fit
GeForce RTX 5090 · 32 GB
Room to spare. 9 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Room to spare. 2.2 GB spare means a 10% error in the size would not change the answer.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first.
Check against your own machine → · Where to rent it hosted →
Or rent it from someone else
Prices checked 2 hours ago — each listing carries its own date.
Io Net, through OpenRouter
Cheapest of the 3 listings we can compare like for like — at 262K of context, out of 6 in the table below. One cheaper row there is outside that comparison: a different quantisation.
- per 1M tokens
- $0.059 in / $0.17 out
- Context served
- 262K
- Throughput
- ~11 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| DeepInfrabf16Direct and through OpenRouter | $0.060 / $0.16checked 2 hours ago | 262K33K max reply through OpenRouter | 155 tok/sthrough OpenRouter | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterNo | DirectUnknownThrough OpenRouterConfirmed |
| OpenRouterOpenRouter's own listing | $0.059 / $0.17checked 2 hours ago | 262K | not measured | Unknown | Unknown | Unknown |
| Io NetThrough OpenRouter | $0.059 / $0.17checked 2 hours ago | 262K131K max reply | 11 tok/s | No | No | Confirmed |
| Darkbloomint4Through OpenRouter | $0.039 / $0.18checked 2 hours ago | 262K33K max reply | 84 tok/s | No | Yesunknown period | Unknown |
| PhalaThrough OpenRouter | $0.070 / $0.20checked 2 hours ago | 262K236K max reply | 178 tok/s | No | No | Confirmed |
| CoreWeavebf16Through OpenRouter | $0.070 / $0.20checked 2 hours ago | 262K236K max reply | 363 tok/s | No | No | Confirmed |
DeepInfra
OpenRouterAcross the 6 listings we hold: 5 say they do not train on prompts (1 of them only through OpenRouter), 0 say they do and 1 does not say. 4 appear in the zero-retention registry we check (1 of them only through OpenRouter); the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| DeepInfrabf16Direct and through OpenRouter | ✓ | ✓ | ✓ |
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| Io NetThrough OpenRouter | ✓ | ✓ | ✓ |
| Darkbloomint4Through OpenRouter | ✗ | ✓ | ✓ |
| PhalaThrough OpenRouter | ✓ | ✓ | ✓ |
| CoreWeavebf16Through OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 5 of 6 listings say yes, 1 says no. JSON output: 6 of 6 listings say yes. Strict schema: 6 of 6 listings say yes.
When we formed this view
Recent changes
What moved
Nemotron 3.5 Lightning moved on 2 DeepInfra listings: DeepInfra through OpenRouter: input −25% ($0.080 → $0.060 per 1M tokens), output −20% ($0.20 → $0.16 per 1M tokens), cache read −25% ($0.040 → $0.030 per 1M tokens); DeepInfra's own listing: input −25% ($0.080 → $0.060 per 1M tokens), output −20% ($0.20 → $0.16 per 1M tokens)What moved
input −40% ($0.065 → $0.039 per 1M tokens)What moved
input −12% ($0.080 → $0.070 per 1M tokens)What moved
input −30% ($0.100 → $0.070 per 1M tokens), output −20% ($0.25 → $0.20 per 1M tokens), cache read −20% ($0.050 → $0.040 per 1M tokens)Each date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- We do not hold the active parameter count for it, so how much of it runs on any one token is unknown to us.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 6 listings does not say whether it trains on prompts, and 1 answers only through OpenRouter, not for its own listing.
- We hold no batch or off-peak rate for any of its listings.
- We hold a decode speed for it, but no prompt-processing (prefill) figure, so how long the input side of a job takes is unknown to us.
Licence and identifiers
What the licence allowsCustom licence, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Custom licence
This model ships custom license terms that don't map to a known template. We haven't parsed them, so commercial use, redistribution and derivatives are unverified — review the original terms before shipping.
Identifiers
- Architecture
- Mixture of experts
- Takes in, gives back
- Text in, text out
- Catalogue slug
- nvidia-nemotron-3-5-lightning