ASUS ProArt GR1X 64GB vs 128GB: Are You Paying for Memory You'll Never Use?

If the largest models you run are about 30 billion parameters or smaller, the 64 GB ASUS ProArt GR1X is enough, and the extra memory won't make anything faster. The 128 GB model earns its premium in three situations:

• You want to run 120B-class models such as gpt-oss-120b.
• You need long context on 70B models.
• You keep several AI tools loaded at once.

Because the memory is soldered and can't be upgraded later, the decision is permanent. Buy for the largest workload you expect over the machine's life, not the one you have today.

Which models fit in 64 GB and 128 GB?

Source: Hugging Face checkpoint sizes; XDA (26 Oct 2026: 64 GB RTX Spark laptop exposed at most 55.4 GB to the GPU); PCWorld (29 Oct 2026: 128 GB ProArt P16 defaults to about half for the GPU, adjustable to 96 GB in MyASUS). The GR1X's own memory-split controls aren't documented yet. Context figures use 16-bit cache sizes; an 8-bit cache halves them.

Does 128 GB make the GR1X faster?

No. Both versions use the same N1X chip with the same memory bandwidth, which ASUS lists as up to 300 GB/s. Bandwidth, not capacity, sets how quickly a model generates text.

XDA showed this on a 64 GB RTX Spark laptop. Qwen3 32B ran at 11.8 tokens per second with 16 GB assigned to the GPU and 12.1 with 48 GB, effectively no difference. If a model fits in 64 GB, it will run at the same speed on the 128 GB model.

Extra memory decides what you can run, not how fast. That's the core of the 128 GB memory trap.

Who is fine with 64 GB?

• Single users running 8B–35B models for chat, writing and coding help through Ollama or LM Studio. Mixture-of-experts models such as Qwen3.6-35B-A3B run well: about 74–78 tokens per second on same-chip laptops in Techtesters' tests (16 Oct 2026). No GR1X benchmark had been published at the time of writing.
• Creators whose main workloads are Adobe apps, ComfyUI image generation with mid-sized models, and 4K video editing.
• Teams using local AI alongside the cloud, sending large or difficult jobs to a cloud service. See whether the GR1X can replace ChatGPT and Claude.

Remember that Windows and your apps need memory too. On 64 GB, roughly 55 GB is the most the GPU will get, and less if you also run WSL2 or large applications.

Who needs 128 GB?

• 120B-class models. gpt-oss-120b and Qwen3.5-122B-A10B are the main reason RTX Spark machines exist. They ran at about 59–63 and 31 tokens per second on 128 GB laptops with the same chip in Techtesters' tests. They don't fit in 64 GB.
• Long documents and conversations. A 70B model at 4-bit plus a 128K-token context needs about 86 GB with a 16-bit cache, or 64 GB with an 8-bit cache.
• Several things loaded at once. An agent setup might keep a chat model, an embedding model and an image model in memory together. Memory adds up quickly.
• Large creative projects. ASUS says the GR1X can render 3D scenes up to 90 GB, which only the 128 GB model can hold.
• Developers using WSL2, who need room for Windows, the Linux environment and GPU memory at the same time. See the Windows on Arm guide for developers.

Memory you'll never use is wasted money, but memory you can't add later is a decision you live with for years. On a soldered machine, size for the biggest model you expect to need.

What does the upgrade cost?

ASUS hasn't announced pricing. SourceIT estimates the GR1X at about S$6,200–7,200 for 64 GB / 1 TB and S$9,000–10,800 for 128 GB / 2 TB, based on comparable RTX Spark devices, not an official ASUS price.

At the midpoints, that's about S$3,200 more for the 128 GB model, which also includes an extra 1 TB of storage. The step matches what we see on other machines. The 64 GB version of ASUS's same-chip ProArt P16 is about two-thirds of the 128 GB version's Singapore price, and the 64 GB and 128 GB versions of other RTX Spark and GB10 systems follow a similar 0.6–0.7 ratio.

A way to frame it: if 128 GB lets you run on the desk even a few tasks a month that you'd otherwise send to a cloud API at heavy volume, or keeps the machine useful for an extra year as models grow, the premium is easy to justify. If you'll stay with 30B-class models, it isn't.

How should you decide?

1. List the largest model you plan to run, its precision and the context length you need.
2. Add memory for anything else that will be loaded at the same time.
3. Compare the total with about 55 GB (64 GB model) or up to about 96 GB (128 GB model with the GPU share raised, based on the ProArt P16).
4. Add headroom for the next two to three years. Open models in the 100–120B class are now common.
5. If the total is well over 96 GB, neither GR1X is the right tool. Look at higher-memory systems.

For model-by-model sizing including context, see choosing GR1X memory based on real needs. For what 70B models feel like in practice, see running a 70B model on the GR1X. Our earlier guide to why 128 GB unified memory matters covers the wider category.

Buying in Singapore

Both GR1X memory options are in our ProArt GR1X collection at price on request. Our figures are SourceIT estimates based on comparable RTX Spark devices, not official ASUS prices. To compare same-chip laptops now, the Lenovo Yoga Pro 9n 64 GB is S$5,897.20, the MSI Prestige N16 Flip AI+ 64 GB is S$6,331.28 and the ASUS ProArt P16 128 GB is S$10,773.20 (listed 8 Nov 2026). Prices change, so ask for a current quotation. Use Request for Quotation on the product page, call (65) 6978 3502, email Sales@sourceit.com.sg or use our contact page.

Frequently asked questions

Is 64 GB enough for the ASUS ProArt GR1X? For models up to about 30–35 billion parameters at 4-bit or 8-bit, yes. It can't run gpt-oss-120b, because Windows exposes only about 55 GB to the GPU on 64 GB RTX Spark machines.

Is the 128 GB GR1X faster? No. Both have the same memory bandwidth, so models that fit in both run at the same speed. The 128 GB model runs larger models and longer contexts.

Can I upgrade the GR1X's memory later? No. The LPDDR5X memory is soldered. Only the SSDs can be changed.

Can a 64 GB GR1X run a 70B model? At 4-bit with a short context, it fits, at about 53 GB with a 32K context. Expect it to be slow, because dense 70B models are limited by memory bandwidth.

How much more does the 128 GB model cost? ASUS hasn't announced pricing. SourceIT estimates about S$3,200 more at the midpoints of our ranges, including the extra 1 TB of storage.