What can you run with 64 GB of memory?

A model does not get the 64 GB. It gets 48 GB on all nine machines here sold with 64 GB, because the system keeps the rest. That holds 27 of the 39 current models at 32k of context, and the strongest of them is Qwen3.8 27B.

Run the numbers on Qwen3.8 27B at 64 GB

Jump to: The machines sold with 64 GB · Models that fit in 64 GB · What 96 GB adds over 64 GB · How much context 64 GB leaves room for · Does a 64 GB machine pay for itself?

The machines sold with 64 GB

There are nine machines here with 64 GB, and each hands a model the same amount. The fourth column is what each one holds at 32k of context, and it is the number that machine's own page prints.

List prices, and the memory each maker publishes. A machine marked previous is one that is no longer sold, priced at what it launched at.

Models that fit in 64 GB

Every current model a 64 GB machine here holds at 32k of context, strongest first. The third column is the whole job at once: the weights plus the cache for 32k of context. The fourth says how many of the machines at this size have room for it, which is the part a size on its own cannot tell you. Speeds are on the Corsair AI Workstation 300, 64GB, the machine this list is cut from; speed follows memory bandwidth rather than memory size, so another machine at 64 GB runs the same model at its own rate, and its page gives every one of them.

For scale, the calculator starts from 80 tok/s for a hosted API and times a local machine against it. 1 of the 27 above reaches it, and the slowest is 7.3 tok/s.

Superseded models are left out of the count, because what a machine is worth buying for is what you would run on it today; the leaderboard ranks every model this site lists, older ones included. Where the cache figure comes from is one section of the memory guide.

What 96 GB adds over 64 GB

The machine here with 96 GB that hands a model the most of it is the RTX PRO 6000 Blackwell, 96GB, at 95 GB, and it holds 33 of the 39. At 64 GB the most is 48 GB, on the Corsair AI Workstation 300, 64GB, holding 27. The step adds Ling 3.0 flash, Qwen3.5 122B-A10B, gpt-oss-120b and Mistral Small 4 (119B-2603), and two more. The strongest either way is Qwen3.8 27B.

The cheapest machine at 96 GB is the Mac Studio M5 Ultra, 96GB at $5,499. The cheapest at 64 GB is the Strix Halo Corsair AI Workstation 300, 64GB at $1,700, so the step costs $3,799.

Both sides are read off the machine at each size that hands a model the most, so the comparison is the best case against the best case. The table above has every machine at 64 GB and what each of them holds.

How much context 64 GB leaves room for

The cache grows with the window you ask for, so the same machine holds fewer models the longer the context. Every machine at this size hands a model 48 GB, so one column covers all of them. A model is counted only where its own context ceiling reaches that far.

Counted over the 39 current models, at the quantisation each is listed at and with the cache at 16 bits. The calculator has a switch for a smaller cache, which changes both what fits and how fast a long window runs.

Does a 64 GB machine pay for itself?

On the Strix Halo Corsair AI Workstation 300, 64GB at $1,700, the cheapest machine at 64 GB with a published price, the model that pays it back soonest at 500k tokens a day is Qwen3.8 27B, in 22 years. At 20M tokens a day, agents running most of the day, it is Qwen3.8 27B in 9.4 months. The Corsair AI Workstation 300, 64GB cannot generate 20M tokens in a day; it manages 14.3M, so that figure is for the most it can do.

The sum is the same everywhere on this site: what the same work costs to rent, less what the electricity costs to generate it, against the price of the machine. The best buys rank the quickest pay-back at every level of use, and what it costs a month puts the machine and the API bill in the same shape.

Put your own usage in

Every figure is at 32k of context unless the row says otherwise, with the cache at 16 bits and each model at the quantisation this site lists it at. Weights are the published file sizes on each model's page, and the cache is worked out from the architecture recorded there. What a model gets is the memory the GPU can address, which each machine's page explains. Speeds say whether anybody measured them; where they were not, they are worked out from memory bandwidth. To change the context, the quantisation or the price you would pay, open the calculator. For the sizes either side of this one, the memory guide has the ladder in full.