How much memory does Qwen3.8-27B need at Q4?

About 17.8 GB at 8,192 tokens: 16.5 GB weights, 0.5 GB KV cache, 0.8 GB overhead. That is an estimate. sourceUrl stays null.

Does Qwen3.8-27B fit on a single 24 GB GPU?

The estimate of 17.8 GB sits under the 23.4 GB we treat as usable on a 24 GB card. It is still an estimate, not a measured file.

Are the tokens per second measured?

No. Speed follows from bandwidth and an assumed utilisation. benchmarkSourceUrl stays null.

model

Qwen3.8-27B: how much memory, and what runs it

Dense vision-language model, Apache-2.0 (the LICENSE file is Apache 2.0, and the card metadata says apache-2.0). The card's parameter headline is 27B. Hugging Face safetensors metadata lists 27,781,427,952 parameters, all BF16. 27.78B is that count to two decimals, so paramsEstimated stays true. 64 layers in text_config: 48 linear_attention (Gated DeltaNet) and 16 full_attention (Gated Attention). 24 query heads, 4 KV heads, head dim 256. Config max_position_embeddings is 262,144. The card says that window is extensible to 1,000,000 tokens with a RoPE change; this catalog uses 262,144. Linear-attention layers are stored with a sliding window of 0 so this formula gives them no token KV. The KV figure uses only the 16 full-attention layers and is an FP16 estimate. The card's citation is August 2026. Hugging Face lists the repo as created on 5 August 2026. That createdAt is repo metadata, not a dated announcement in the citation. Vision projector and the MTP drafter are separate files, not rows in the memory table. On unsloth/Qwen3.8-27B-GGUF, checked 19 Oct 2026: mmproj-F16 is 0.93 GB (927,607,488 bytes) and mmproj-BF16 is 0.93 GB (931,146,432 bytes). MTP/mtp-Qwen3.8-27B-Q4_0.gguf is 1.37 GB (1,369,590,656 bytes). Main GGUF sizes are decimal GB from that tree, single files: UD-IQ2_XXS 7.27, UD-IQ3_XXS 10.93, UD-IQ4_XS 14.25, UD-Q4_K_M 16.46, UD-Q4_K_XL 17.56, UD-Q5_K_XL 20.88, UD-Q6_K 21.98, Q8_0 29.05. A separate community speed table for one RTX 3090 is stored on this row. It is not the calculator estimate.

Qwen3.8-27B needs about 17.8 GB at Q4_K_M with 8K context (estimate). See which GPUs and Macs fit it, estimated tokens/sec, and EU prices.

Quantisation	Estimated memory
Q4	17.8 GBestimate
Q5	21.1 GBestimate
Q8	31.1 GBestimate
FP16	56.9 GBestimate

At 8,192 tokens of context. Catalog quant rates are a formula. sourceUrl on that memory stays null.

Parameter count source: huggingface.co/Qwen/Qwen3.8-27B. Marked estimated because the published count is rounded.

community measurement, not ours

Canonical single-GPU figures on that results page: mean of five runs, one RTX 3090, the other replica stopped, the ~1K-token prompt column, 512-token greedy completions, reasoning disabled, 250 W power cap. The page says every rung is the same 4-bit checkpoint. llama.cpp out of the box 27.6 tokens per second; llama.cpp tuned (FA + q8 KV + MTP) 42.6; stock vLLM + MTP k=4 85.6; patched vLLM + MTP k=4 167.8; patched vLLM + DFlash2 k=7 212.1. The page dates the single-replica correction 31 August. results.html was last committed on 18 September 2026, which is the published date stored here. Not a localLLMrig measurement. Calculator estimates keep benchmarkSourceUrl null.

Hardware	Tokens per second
RTX 3090, llama.cpp out of the box	27.6community measurement, not ours
RTX 3090, llama.cpp tuned (FA + q8 KV + MTP)	42.6community measurement, not ours
RTX 3090, stock vLLM + MTP k=4	85.6community measurement, not ours
RTX 3090, patched vLLM + MTP k=4	167.8community measurement, not ours
RTX 3090, patched vLLM + DFlash2 k=7	212.1community measurement, not ours

Qwen3.8-27B on RTX 3090, published 2026-09-18

VRAM per quant and the cheapest rig for Qwen3.8-27B

Open in the calculator

Three setups

Qwen3.8-27B · Q4 · 8,192 ctx

Estimated memory

17.8 GBestimate

16.5 GB weights · 0.5 GB KV cache · 0.8 GB overhead

01 · Fastest
PC with RTX 5090 32GB
Highest bandwidth in a single card. 32 GB is plenty for 32B, still too tight for 70B Q4.
$8,283€7,399
price as of 16 Oct 2026
64 GB DDR4 (estimate): $672€600. From the 16 Oct 2026 fact-check.
rest of system (estimate): $560€500. The total is the GPU price plus the RAM line plus this estimate. It is not a checked system price.
price unverified
GPU line is the new in-stock floor €6,299 from PriceSquirrel at Proshop (FI) (https://www.pricesquirrel.com/gpus/rtx-5090). 64 GB DDR4 (estimate) €600, from the 16 Oct 2026 fact-check. Rest of system (estimate) €500. Other parts from €500 in that fact-check. Not a shop listing. The parts list names 96 GB DDR5. That kit was not priced, so this total uses the sourced 64 GB DDR4 line. The total is the GPU price plus the RAM line plus that rest estimate. It is not a checked system price.
tok/s estimate
81
estimate
needed
17.8 GB
full fit
usable
31.4 GB
VRAM
Why this one
- The model fits entirely in VRAM or unified memory. No CPU offload.
- High memory bandwidth, so a higher tokens-per-second estimate.
- Highest estimated speed in this shortlist.
Parts
- GeForce RTX 5090 32GB
- 96 GB DDR5 is named here and was not priced.
- 64 GB DDR4 (estimate): €600. from the 16 Oct 2026 fact-check.
- Rest of system (estimate): €500. Other parts from €500 in that fact-check. Not a shop listing.
- Ryzen 7 9700X or similar
- 1200W PSU, roomy case
EU shops
US shops are not linked yet.
02 · Best value
Used PC with RTX 3090 24GB
Best single card here for 32B Q4. 70B Q4 does not fully fit.
$2,602€2,324
price as of 16 Oct 2026
64 GB DDR4 (estimate): $672€600. From the 16 Oct 2026 fact-check.
rest of system (estimate): $560€500. The total is the GPU price plus the RAM line plus this estimate. It is not a checked system price.
price unverified
GPU line is the used median €1,224, labelled used, from pcprice.watch (https://www.pcprice.watch/gpu-buying-guide/rtx-3090-price-used-and-specs). 64 GB DDR4 (estimate) €600, from the 16 Oct 2026 fact-check. Rest of system (estimate) €500. Other parts from €500 in that fact-check. Not a shop listing. The total is the GPU price plus the RAM line plus that rest estimate. It is not a checked system price.
used
tok/s estimate
43
estimate
needed
17.8 GB
full fit
usable
23.4 GB
VRAM
Why this one
- The model fits entirely in VRAM or unified memory. No CPU offload.
- High memory bandwidth, so a higher tokens-per-second estimate.
- Second-hand or a previous generation. The price moves around.
- Most estimated tokens per euro in this shortlist.
Parts
- GeForce RTX 3090 24GB, used
- 64 GB DDR4 (estimate): €600. from the 16 Oct 2026 fact-check.
- Rest of system (estimate): €500. Other parts from €500 in that fact-check. Not a shop listing.
- Ryzen 7 5700X or similar
- 850W PSU, decent airflow
EU shops
US shops are not linked yet.
03 · Other platform
Mac Studio M5 Ultra 96GB
High bandwidth and enough memory for 70B Q4/Q5. FP16 70B does not fit.
$7,444€6,649
price as of 16 Oct 2026
Belgian Apple Store
Belgian Apple Store price for this config, from the 16 Oct 2026 fact-check.
tok/s estimate
34
estimate
needed
17.8 GB
full fit
usable
72.0 GB
unified
Why this one
- The model fits entirely in VRAM or unified memory. No CPU offload.
- Memory left for a longer context or a heavier quantisation.
- Unified memory: weights and KV cache share one pool.
- High memory bandwidth, so a higher tokens-per-second estimate.
- One box, little noise, no discrete GPU.
- A different brand, so you don't see the same PC three times.
Parts
- Mac Studio M5 Ultra, 96 GB, 1.2 TB/s
EU shops
US shops are not linked yet.

With CPU offload, and therefore slower, this could also work: Mid-range PC with RTX 4060 Ti 16GB, ~8.0 tok/s estimate, $2,238€1,999.

Tokens per second are estimated from memory bandwidth, not from a measured benchmark. Prices incl. VAT, excl. shipping. USD = FX estimate, excl. sales tax. A price without a checked listing is marked price unverified. Run llama-bench yourself before you buy.

Want a build tailored to you? Custom report

Get new builds when GPU prices move

Embed this

Pair

[![Qwen3.8-27B Q4 on RTX 3060 12GB: ✗ · needs 17.8 GB (estimate)](https://localllmrig.com/badge/qwen3.8-27b/rtx-3060-12gb.svg)](https://localllmrig.com/can-i-run/qwen3.8-27b/rtx-3060-12gb?utm_source=badge)

Model badge

[![Qwen3.8-27B Q4: needs ~17.8 GB VRAM (estimate)](https://localllmrig.com/badge/qwen3.8-27b.svg)](https://localllmrig.com/models/qwen3.8-27b?utm_source=badge)

Common questions

How much memory does Qwen3.8-27B need at Q4?: About 17.8 GB at 8,192 tokens: 16.5 GB weights, 0.5 GB KV cache, 0.8 GB overhead. That is an estimate. sourceUrl stays null.
Does Qwen3.8-27B fit on a single 24 GB GPU?: The estimate of 17.8 GB sits under the 23.4 GB we treat as usable on a 24 GB card. It is still an estimate, not a measured file.
Are the tokens per second measured?: No. Speed follows from bandwidth and an assumed utilisation. benchmarkSourceUrl stays null.