What hardware do you need to run Nemotron 3.5 Lightning 30B-A3B?

Nemotron 3.5 Lightning 30B-A3B at Q4_K_M is 25 GB of weights, with a context ceiling of 256k tokens. NVIDIA's Mamba hybrid: almost no KV cache, so long contexts cost nothing in memory. On the cheapest machine that runs it, a Radeon AI PRO R9700, 32GB, it generates 127 tok/s at 32k of context, worked out from memory bandwidth.

How good is Nemotron 3.5 Lightning 30B-A3B, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 14, which puts it in the Below every hosted tier band. Fine for simple, well-specified tasks. Noticeably less capable than anything the big labs sell today. See the whole table.

Head to head with its neighbours on the leaderboard: vs Qwen3.5 9B above it, vs Qwen3 235B-A22B Instruct 2507 below. One more model needs much the same memory at 32k of context: vs Granite 4.2 30B.

What Nemotron 3.5 Lightning 30B-A3B costs either way

Renting the same model costs $0.025 per million input tokens and $0.12 per million output (OpenRouter, cheapest active endpoint, checked 2026-10-29). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill. What a million tokens costs each way puts the two prices side by side.

Machines that run Nemotron 3.5 Lightning 30B-A3B

Every machine here runs it, but not to the same length. The Mac Studio M5 Max, 36GB stops at 128k; the AMD Radeon AI PRO R9700, 32GB takes it to 256k, which is this model's own ceiling and the longest the calculator offers. The weights are the same size on every machine; what differs is the memory left for the key-value cache, which grows with every token you keep.

The table is the cheapest machine in each family, not the only one that runs it: 38 of the 56 machines on this site hold it at 32k. Within a family only memory decides, so that is every Mac Studio at any size, every NVIDIA card with 32GB or more, every Mac mini and MacBook Pro with 48GB or more, every Strix Halo box with 64GB or more, the Radeon AI PRO R9700, 32GB and the DGX Spark, 128GB. One table sets every machine against the others: price, the memory its GPU can use, what it holds and how long it takes to pay for itself.

One machine per family, cheapest first. Speeds are measured where a public benchmark exists and estimated from memory bandwidth otherwise; the calculator says which for any configuration. The longest context is the longest setting the calculator offers that the machine still holds this model at, cache included, and no machine is shown taking it past its own 256k limit. Each one opens the calculator on that machine at that length.

For scale, the calculator starts from 80 tok/s for a hosted API and times a local machine against it. 2 of the 7 above reach it, and the slowest is 35 tok/s.

The cheapest machine that runs it is sold with 32 GB. There is a page on what 32 GB runs, machine by machine: it names the 6 machines here sold at that size, and the 32 current models the roomiest of them holds at 32k, this one among them.

Run Nemotron 3.5 Lightning 30B-A3B at 256k on the AMD Radeon AI PRO R9700, 32GB

How much memory does Nemotron 3.5 Lightning 30B-A3B need at 256k?

Every figure above is priced at 32k of context. At 256k, which is this model's own ceiling and the longest the calculator offers, Nemotron 3.5 Lightning 30B-A3B needs 27.1 GB rather than 25.7 GB. The weights are the same either way; the key-value cache goes from 0.2 GB to 1.6 GB.

27 of the 37 machines here still sold at a published price hold Nemotron 3.5 Lightning 30B-A3B at 32k and 26 hold it at 256k. The cheapest still sold is the same at both lengths: the Radeon AI PRO R9700, 32GB at $1,299, card only. The Radeon AI PRO R9700, 32GB is priced as the card alone, so add the PC around it before comparing it with a complete computer.

On the Radeon AI PRO R9700, 32GB it runs 127 tok/s estimated at 32k and 83 tok/s estimated at 256k. Every token generated reads the whole cache back, so a longer window is slower as well as bigger.

The specifics

Parameters

31.6B, of which 3B are active per token

Quantisation

Q4_K_M

Weights on disk

25 GB

KV cache

0.2 GB at 32k context — Only 6 of 52 layers are attention at all; 23 are Mamba blocks with a fixed-size state and 23 are feed-forward only. The context is close to free. How weights and cache add up.

Maximum context

256k tokens

Licence

OpenMDW 1.1

Sources

nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16