Home›Hardware›PC with an RTX 4060 8 GB
PC with an RTX 4060 8 GB
NVIDIA graphics card8 GiB272 GB/s15.0 TFLOPS
Once the system takes its share, this machine leaves 7.20 GiB for the model: room for 7 of the 26 models we measure at the reference compression, and the heaviest one that fits is qwen3.5-9b. And how fast does it answer? That’s not down to the chip, but to its 272 GB/s of memory bandwidth.
Left for the model
Estimated
Memory bandwidth
Estimated
Models that fit
Measured by Local AI Scope
Videos: PC with an RTX 4060 8 GB 2
The best walk-throughs we found. Press play and it opens right here.
The videos live on YouTube, and nothing loads from there until you press play (no-cookie mode). The thumbnail is served by us.
The memory you can actually use
This is why the number on the box is not the number you get. The desktop, the compositor and the graphics context never give their share back, and a model that does not leave them room does not load — it crashes.
| Memory | GiB |
|---|---|
| Memory on the box | 8 |
| Reserved for the system | −0.80 |
| Left for the model | 7.20 |
Memory bandwidth
Two different limits, and they are easy to mix up. While the model is writing its answer, bandwidth rules: every token means reading the weights it uses again — all of them in a dense model; in a mixture of experts, the shared layers plus only the active experts — so tokens per second is roughly bandwidth divided by that amount. While the model is reading your document, compute rules: the whole input is processed at once. That is why a laptop with no GPU can take minutes before the first word appears and then type at a tolerable pace.
272 GB/s — Derived, not published by the manufacturer · the manufacturer publishes the bus width but neither the memory speed nor the bandwidth. This figure is the usual arithmetic derivation (17 Gbps x 128 bit / 8 = 272 GB/s) and we could not confirm it at source (Source).
Compute: 15.0 TFLOPS (FP32 shader TFLOPS) — Vendor declared, under a different name · the manufacturer publishes this number as FP32 shader TFLOPS, not as FP16; checked against the manufacturer’s page on 2026-08-25 (Source). The figures in this line do not all come from the same measure across machines, because each manufacturer publishes a different one. Compare them between machines only with that in mind.
Which models fit here, and which do not
Fitting means three things at once fit in the memory left over: the model file, its context memory, and the runtime’s working headroom. Context memory is computed from each model’s declared attention pattern layer by layer, not from the generic formula — which is why several models fit here that the generic formula would rule out.
| Model | Compression shown | Model size | 8k | 32k | 128k | Largest context that fits |
|---|---|---|---|---|---|---|
| qwen3-1.7b | Q4_K_M |
1.03 GiB | yes | yes | — | 32k |
| qwen3-4b-2507 | Q4_K_M |
2.33 GiB | yes | no | no | 16k |
| gemma4-e2b | Q4_K_M |
2.89 GiB | yes | yes | yes | 128k |
| gemma4-e4b | Q4_K_M |
4.63 GiB | yes | yes | no | 32k |
| qwen3-8b | Q4_K_M |
4.68 GiB | yes | no | — | 8k |
| granite-4.2-8b | Q4_K_M |
4.98 GiB | yes | no | no | 8k |
| qwen3.5-9b | Q4_K_M |
5.29 GiB | yes | no | no | 16k |
| gemma4-12b | Q4_K_M |
6.63 GiB | no | no | no | — |
| gpt-oss-20b | Q4_K_M |
10.83 GiB | no | no | no | — |
| mistral-small-24b | Q4_K_M |
13.35 GiB | no | no | no | — |
| qwen3.8-27b | Q4_K_M |
15.33 GiB | no | no | no | — |
| muse-glimmer-30b | Q4_K_M |
15.61 GiB | no | no | no | — |
| qwen3.6-27b | Q4_K_M |
15.66 GiB | no | no | no | — |
| gemma4-26b-a4b | Q4_K_M |
15.78 GiB | no | no | no | — |
| gemma4-31b | Q4_K_M |
17.07 GiB | no | no | no | — |
| qwen3-coder-30b | Q4_K_M |
17.28 GiB | no | no | no | — |
| qwen3.6-35b-a3b | Q4_K_M |
20.61 GiB | no | no | no | — |
| flint-40b | Q4_K_M |
23.61 GiB | no | no | no | — |
| nemotron-3.5-lightning-30b | Q4_K_M |
23.73 GiB | no | no | no | — |
| gpt-oss-120b | Q4_K_M |
58.46 GiB | no | no | no | — |
| llama4-scout-17b | Q4_K_M |
60.87 GiB | no | no | no | — |
| nemotron-3.5-super-120b | Q4_K_M |
65.29 GiB | no | no | no | — |
| qwen3.8-flash-next | IQ4_XS |
87.25 GiB | no | no | no | — |
| deepseek-v4-flash | IQ4_XS |
127.28 GiB | no | no | no | — |
| glm-5.3-flash | Q4_K_M |
187.04 GiB | no | no | no | — |
| glm-5.2 | Q4_K_M |
433.83 GiB | no | no | no | — |
Sizes are the real published files at the reference compression, one step per row: Q4_K_M where the author publishes it, the nearest neighbour where they do not. Every row states which one it is showing. Q4_K_M is our quality floor: below it the loss is audible in the answers. A model that would only fit here at a harsher compression is listed as not fitting, on purpose. A dash in a context column means the model itself does not offer that context, so there is nothing to fit.
Models that fit — 7 of the 26 models measured
- qwen3-1.7b
Q4_K_M— context 32k tokens - qwen3-4b-2507
Q4_K_M— context 16k tokens - gemma4-e2b
Q4_K_M— context 128k tokens - gemma4-e4b
Q4_K_M— context 32k tokens - qwen3-8b
Q4_K_M— context 8k tokens - granite-4.2-8b
Q4_K_M— context 8k tokens - qwen3.5-9b
Q4_K_M— context 16k tokens
Models that do not fit — 19 of the 26 models measured
- gemma4-12b
Q4_K_M— short by 0.48 GiB - gpt-oss-20b
Q4_K_M— short by 4.40 GiB - mistral-small-24b
Q4_K_M— short by 7.45 GiB - qwen3.8-27b
Q4_K_M— short by 9.06 GiB - muse-glimmer-30b
Q4_K_M— short by 9.21 GiB - qwen3.6-27b
Q4_K_M— short by 9.39 GiB - gemma4-26b-a4b
Q4_K_M— short by 9.53 GiB - gemma4-31b
Q4_K_M— short by 11.63 GiB - qwen3-coder-30b
Q4_K_M— short by 11.13 GiB - qwen3.6-35b-a3b
Q4_K_M— short by 14.17 GiB - flint-40b
Q4_K_M— short by 17.22 GiB - nemotron-3.5-lightning-30b
Q4_K_M— short by 17.23 GiB - gpt-oss-120b
Q4_K_M— short by 52.08 GiB - llama4-scout-17b
Q4_K_M— short by 55.10 GiB - nemotron-3.5-super-120b
Q4_K_M— short by 58.80 GiB - qwen3.8-flash-next
IQ4_XS— short by 80.82 GiB - deepseek-v4-flash
IQ4_XS— short by 121.09 GiB - glm-5.3-flash
Q4_K_M— short by 180.60 GiB - glm-5.2
Q4_K_M— short by 427.92 GiB
Willing to compress a bit more? Then not everything on this list is out of reach. In the 7.20 GiB of VRAM left for the model, gemma4-12b runs at IQ4_XS (5.94 GiB) with up to 8k tokens of context. Below the compression in the table a model loses quality; that is why our count sticks to the table.
The machines on either side
- PC with an RTX 3060 12 GB — 11.20 GiB left for the model, 8 / 26 models
- PC with an RTX 4070 12 GB — 11.20 GiB left for the model, 8 / 26 models
- Mac with M4 and 16 GB unified memory — 13.00 GiB left for the model, 9 / 26 models
- Mac mini with M6 and 16 GB — 13.00 GiB left for the model, 9 / 26 models
Price and availability
What a machine costs decides as much as what fits inside it, so the price belongs on this page. What follows is only what the manufacturer publishes, with the market it applies to and the day it was read.
There is no price: this row is a class of machine, not a single product with a price list.
What we earn on this page
Nothing, today. There is no affiliate link on this page and no commercial agreement behind any figure on it. If a buying link ever appears here, it will be marked as such and the commission declared in this same block. Two rules do not change on that day: the ranking on this site is computed before commercial availability is applied, and «best» never means «pays most».