Qwen3-32B on RTX 5080: Does not fit (Sep 2026)

Conditional memory judgement for one configuration. Change context, quant, or engine and the conclusion can change. Configuration used on this page: GGUF Q4_K_M, 8,192 cached tokens, 1 sequence, engine llama.cpp. Status: Insufficient memory. Software: Not verified..

Insufficient memoryNot verifiedThis result is an estimate, not a measured trace.

Verdict on GeForce RTX 5080 16GB (desktop)

No. The smallest published GGUF (Q4_K_M) needs 20.5 GiB–21.4 GiB against 14.0 GiB of usable VRAM (after the driver reserve). It would only run by spilling layers to system RAM, which this site does not model.

Published GGUF quants for Qwen3-32B on GeForce RTX 5080 16GB (desktop) at 8,192 tokens
GGUF quantFile weightsKV cacheEstimated totalVerdict
Q8_032.4 GiB2.0 GiB34.6 GiB–35.7 GiBDoes not fit
Q6_K25.0 GiB2.0 GiB27.2 GiB–28.2 GiBDoes not fit
Q5_K_M21.6 GiB2.0 GiB23.7 GiB–24.7 GiBDoes not fit
Q4_K_M18.4 GiB2.0 GiB20.5 GiB–21.4 GiBDoes not fit

Engine llama.cpp, one sequence, fp16 KV. File weights are the published GGUF sizes in the catalog. Totals come from the same calculator that powers the tool below; they are estimates, not a measurement of this setup.

Fixed page configuration

No published GGUF fits: even GGUF Q4_K_M needs 20.5 GiB-21.4 GiB against 14.00 GiB net.

Technical identifiers and share data
  • Modelqwen3-32b
  • Hardwarenvidia-geforce-rtx-5080-16gb-desktop
  • Quantizationgguf-q4_k_m
  • Share fragment for this identityv2.eyJzY2hlbWFWZXJzaW9uIjoyLCJjYWxjdWxhdG9yVmVyc2lvbiI6IjAuMy4wIiwiZGF0YXNldFZlcnNpb24iOiIyMDI2LTEwLTAyLjEiLCJjb25maWciOnsibW9kZWxJZCI6InF3ZW4zLTMyYiIsImhhcmR3YXJlSWQiOiJudmlkaWEtZ2Vmb3JjZS1ydHgtNTA4MC0xNmdiLWRlc2t0b3AiLCJkZXZpY2VDb3VudCI6MSwicXVhbnRpemF0aW9uSWQiOiJnZ3VmLXE0X2tfbSIsImNvbnRleHRUb2tlbnMiOjgxOTIsImNvbmN1cnJlbnRTZXF1ZW5jZXMiOjEsImVuZ2luZSI6ImxsYW1hLmNwcCIsImt2VHlwZSI6ImZwMTYifX0

Memory breakdown

Weight artifact
18.40 GiB (19,762,149,024 bytes)
Raw KV cache
2.00 GiB (2,147,483,648 bytes)
Engine extras / workspace
0.13 GiB–1.01 GiB
Total estimated demand
20.53 GiB–21.41 GiB
Net calculator budget
14.00 GiB
OS / display reserve
2.00 GiB
Remaining vs net budget
-7.41 GiB–-6.53 GiB

21.41 GiB high estimate vs 14.00 GiB net

Enough memory does not mean this engine, quant, and OS will run. Support does not mean it will be fast.

Other published files on this device

Same model, device and context (8,192 tokens, 1 sequence, FP16 KV). Each row is one published weight-file set with its usual engine, sorted by memory.

QuantizationEngineEstimated memoryFit
BF16 safetensorsvllm63.65 GiB–65.58 GiBDoes not fit
GGUF Q8_0 (Qwen)llama.cpp34.55 GiB–35.72 GiBDoes not fit
GGUF Q6_K (Qwen)llama.cpp27.16 GiB–28.18 GiBDoes not fit
GGUF Q5_K_M (Qwen)llama.cpp23.75 GiB–24.69 GiBDoes not fit
GGUF Q4_K_M This pagellama.cpp20.53 GiB–21.41 GiBDoes not fit

Assumptions that affect this result

Open this combo in the calculatorKV Cache

Related pages

Sources

Calculator

This combo page keeps the route model and device. Customize to change them on the main calculator; the share then keeps that identity.

Loading catalog snapshot…

Choose a model first, then check a device.

Quantization

Device type

Workload

Context length primarily affects KV cache memory.

Advanced settings
Runtime

Engine and KV cache type for this estimate.

Workload

Overhead scenario and the prefill workspace.

Topology

How the model is split across devices. Memory is not added together.

Replica count is an advanced deployment option. Quantity is nodes or GPUs, not replicas.

Budget

System reserve and safety margin. An empty number keeps the previous value.

GiB (2^30 bytes)

%

Custom model

Import a config or type the architecture. Nothing is scanned until you import.

Parsed in the browser. The file is not uploaded.

Only huggingface.co, config.json, no token. Private repos are rejected.

Search or pick a catalog checkpoint to see memory demand.

Search or pick a catalog checkpoint to see memory demand.