onrup

Qwen3 · Alibaba

Fine-tuning Qwen3 14B

What does it take to fine-tune Qwen3 14B?

Qwen3 14B needs 28 GB for half-precision LoRA training and 20 GB in four-bit, and 26 GB to serve. It supports supervised fine-tuning, preference tuning (dpo), reinforcement learning (grpo), under the Apache 2.0 licence. The cheapest qualifying class is RTX 4000 Ada at $0.09 per GPU-hour.

The first rung where a 48GB class becomes necessary for half-precision LoRA. Worth the step from 8B when the task needs to hold more of the world in its head at once — long documents, dense domain knowledge, multi-hop reasoning — and not otherwise.

What to know before choosing it

The reason to step up from 8B is not general quality, which improves less than the parameter jump suggests. It is working capacity: holding a long document, a dense domain vocabulary and a multi-hop question at the same time. If your task does not need all three at once, 8B will usually match this at half the memory.

Twenty-eight gigabytes for half-precision LoRA is the awkward number — no 24GB class covers it, so this is where you either move to a 48GB class or accept quantised training. Compare the total run cost of both before assuming the cheaper card wins.

Specification

Qwen3 14B specification
Parameters14B
ArchitectureDense
Size tierMid-large
LoRA training memory28 GBHalf precision, frozen base
QLoRA training memory20 GBFour-bit base, higher-precision adapter
Serving memory26 GBHalf precision, before attention cache
ObjectivesSFT, DPO, GRPO
LicenceApache 2.0
RepositoryQwen/Qwen3-14B

What it costs to train

Every class with enough memory for four-bit training, cheapest first. Total cost is the rate multiplied by wall time, so the cheapest rate is not always the cheapest run — a faster class that finishes sooner frequently wins.

GPU classVRAMTraining / hrServing / hrFits
RTX 4000 Ada20 GB$0.09$0.11QLoRA only
L424 GB$0.17$0.20QLoRA only
RTX 309024 GB$0.22$0.26QLoRA only
RTX 409024 GB$0.38$0.43QLoRA only
A4048 GB$0.42$0.48LoRA and QLoRA
RTX 6000 Ada48 GB$0.61$0.71LoRA and QLoRA
A600048 GB$0.65$0.75LoRA and QLoRA
L40S48 GB$0.78$0.90LoRA and QLoRA
A100 40 GB40 GB$1.17$1.35LoRA and QLoRA
A100 80 GB80 GB$1.68$1.94LoRA and QLoRA
H100 80 GB80 GB$2.59$2.99LoRA and QLoRA
H200141 GB$4.55$5.25LoRA and QLoRA

Half-precision LoRA needs 28 GB, so the cheapest class for it is A40 at $0.42 per hour. Below that, training has to be quantised.

Serving fits on A40 at $0.48 per hour — before the attention cache, which grows with context length and concurrency.

Good starting point for

Preference tuning on this model

Preference tuning holds a frozen reference copy of the model alongside the one being trained, so budget roughly 56 GB rather than 28 GB. That is the thing that catches people out — supervised training on this model fits on a class that preference tuning will overflow.

When preference tuning beats supervised fine-tuning →

Other sizes in this family

ModelParamsQLoRAObjectives
Qwen3 0.6B600M3 GBSFT, DPO, GRPO
Qwen3 1.7B1.7B4 GBSFT, DPO, GRPO
Qwen3 4B4B8 GBSFT, DPO, GRPO
Qwen3 8B8B14 GBSFT, DPO, GRPO
Qwen3 30B-A3B30B36 GBSFT
Qwen3 32B32B48 GBSFT

Comparable sizes elsewhere

Frequently asked questions

How much VRAM does it take to fine-tune Qwen3 14B?

28 GB for half-precision LoRA and 20 GB for four-bit QLoRA. Serving needs 26 GB. Preference tuning roughly doubles the training figure, because a frozen reference model is held alongside the one being trained.

What is the cheapest way to fine-tune Qwen3 14B?

Four-bit QLoRA on RTX 4000 Ada at $0.09 per GPU-hour is the cheapest class that meets the 20 GB threshold. Note that quantised training is slower per step, so a faster class sometimes costs less over the whole run.

Can I download the weights after fine-tuning Qwen3 14B?

Yes. Every finished run exposes its trained weights for download, and publishing to a model hub is a single call with a generated model card recording the base model and version the adapter applies to.

What licence does Qwen3 14B carry?

Apache 2.0. The licence follows the fine-tune — a derivative inherits the base model’s terms, and those terms pass to anyone you give the model to.

Last verified 6 August 2026. Memory thresholds are the platform's own admission limits.

Fine-tune Qwen3 14B

Upload a dataset, forecast the run, and see the cost before any compute is leased.