More
Best rig by budgetRent before you buySubmit a buildNewsletterFree datasetAdvertise
Submit build

Builds / RTX 5080 Laptop 16GB

RTX 5080 Laptop 16 GB + 32 GB RAM (LM Studio, MTP): Qwen3.6 35B-A3B at 102 tok/s @17k context after offload retune

Last updated:

Source: dev.to ↗📏 How measured: reported by the original source; RigBase re-checked the numbers at that source
GPU(s)
RTX 5080 Laptop 16GB
Amazon · eBay (used) · Newegg
VRAM / unified memory (GB)
16
CPU
not stated
Motherboard
gaming laptop
Amazon · eBay (used) · Newegg
RAM
32 GB
Amazon · eBay (used) · Newegg
PSU
not stated
Other
LM Studio serving over the network; model 21.7 GB so experts partially in system RAM. After retune: GPU Offload max (all 40 layers on GPU), force MoE weights onto CPU for 40% of layers (~16), KV cache q8_0 (~14.5 GB VRAM in use). Speculative decoding (MTP) enabled via a separate load flag
Reported total cost
not stated
Power
not stated

Measured speed (generation tokens/sec)

ModelQuantEnginetok/sPrompt processing (tok/s)Details
Qwen3.6 35B-A3B4-bit (LM Studio / MoE)LM Studio (CUDA), MTP on102not stated⚡ spec-decode (MTP)context: ~17k tokens of agent context after retune📏 Author's before/after table: at 17k context, generation rose from 21 tok/s to 102 tok/s after the offload retune; at 42k from 9.5 to 94 tok/s. Short-context speed after retune was not re-measured (before retune: 54 tok/s short). MTP draft head on — without the flag the author says the model runs about twice as slow

dev.to post of 4 Nov 2026. Laptop RTX 5080 (16 GB), not a desktop 5080. Headline is the 17k-context figure after the MoE/CPU-expert retune with MTP on; not comparable with short-context plain-decode rows. Agent quality scores from the same post are not tok/s and are omitted.

Current GPU street price (estimate, not the builder's cost)Estimated GPU cost for this build today: ≈€1,770
New retail prices in Germany incl. VAT; used cards are often much cheaper. Prices change daily.
Source: https://dev.to/ngirchev/can-local-llms-replace-opus-and-codex-i-tested-a-macbook-an-rtx-5080-a-dgx-spark-and-cheap-4hdb

Blank fields = not stated in the source. We never estimate build data; GPU street prices above are separately sourced and dated.

Embed this result

Copy into a README or blog. Links back to this build. See badge docs (CC BY 4.0 attribution). The SVG does not invent speeds — numbers stay on the build page.

Markdown

[![Measured on RigBase](https://getrigbase.com/badge/rtx-5080-laptop-16gb-qwen36-35b-a3b-lmstudio-mtp-ngirchev.svg)](https://getrigbase.com/builds/rtx-5080-laptop-16gb-qwen36-35b-a3b-lmstudio-mtp-ngirchev/)

HTML

<a href="https://getrigbase.com/builds/rtx-5080-laptop-16gb-qwen36-35b-a3b-lmstudio-mtp-ngirchev/" rel="noopener"><img src="https://getrigbase.com/badge/rtx-5080-laptop-16gb-qwen36-35b-a3b-lmstudio-mtp-ngirchev.svg" alt="Measured on RigBase" height="20"/></a>

Preview: Measured on RigBase · How to use

Rent before you buy: try this model on a cloud GPU for a few dollars first: Runpod (referral link; sign up with Google to get the new-user bonus)

Quick answers: Can RTX 5080 run Qwen3.6 35B-A3B?

Speculative-decoding rows on this page are not like-for-like with plain decode. See how much MTP and DFlash really speed things up (measured pairs).