What can you run with 192 GB of memory?

A model does not get the 192 GB. It gets 160 GB on the one machine here sold with 192 GB, because the system keeps the rest. That holds 45 of the 51 current models at 32k of context, and the strongest of them is Qwen3.8 Flash Next.

Run the numbers on Qwen3.8 Flash Next at 192 GB

Jump to: The machines sold with 192 GB · Models that fit in 192 GB · What 256 GB adds over 192 GB · How much context 192 GB leaves room for · Does a 192 GB machine pay for itself?

The machines sold with 192 GB

One machine here comes with 192 GB, and the fourth column is what it holds at 32k of context: the number its 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 192 GB

Every current model the one 192 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 Framework Desktop, 192GB, the machine this list is cut from; speed follows memory bandwidth rather than memory size, so another machine at 192 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. 4 of the 45 above reach it, and the slowest is 2.4 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Superseded models are left out of the count, because what a machine is worth buying for is what you would run on it today; the leaderboard ranks every model this site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 256 GB adds over 192 GB

The machine here with 256 GB that hands a model the most of it is the Mac Studio M5 Ultra, 256GB, at 192 GB, and it holds 47 of the 51. At 192 GB the most is 160 GB, on the Framework Desktop, 192GB, holding 45. The step adds GLM-5.3-Flash and Inkling Small. The strongest goes from Qwen3.8 Flash Next to GLM-5.3-Flash.

The step buys speed as well as room. The Mac Studio M5 Ultra, 256GB reads memory at 1200 GB/s and the Framework Desktop, 192GB at 273 GB/s, so Qwen3.8 Flash Next, the strongest model both of them hold, runs at 74 tok/s on the 256 GB machine and 31 tok/s on the 192 GB one, both worked out from memory bandwidth.

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 for room, which is not the same as the best case for speed. The table above has every machine at 192 GB and what each of them holds.

How much context 192 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 160 GB, so one column covers all of them. A model is counted only where its own context ceiling reaches that far.

Counted over the 51 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 192 GB machine pay for itself?

No machine here with 192 GB has a published price, so this page cannot say what one pays back. It can say what one would have to cost. At 500k tokens a day the model that saves the most of the 45 the Framework Desktop, 192GB holds is Ornith 1.5 122B, at $0.18 a day against renting the same work, so for the money to come back it would have to cost no more than $65 inside one year, $130 inside two years and $195 inside three years. At 20M tokens a day, agents running most of the day, it is Ornith 1.5 122B at $7.12 a day, and those figures become $2,599, $5,198 and $7,797. Those are prices to beat, not prices anyone has announced. The 133 W the electricity is priced at is a stand-in: nobody has put a meter on the Framework Desktop, 192GB, and its own page says what the figure borrows. The calculator takes yours: put what you would pay beside the machine and it prices the rest.

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. The sizes either side of this one are 128 GB and 256 GB. The memory guide has the ladder in full.