Qwen3 14B on an L40S
Yes, you can fine-tune Qwen3 14B on one L40S (48 GB) with QLoRA or LoRA. 4-bit inference: yes, 8.3 GB of weights, room for 128K tokens of context or 51 concurrent 4K-token sequences.
Verdict by method
| Fine-tuning | Memory at 2K tokens | One L40S? | Longest sequence (batch 1) | Largest batch at 2K | Largest LoRA rank |
|---|---|---|---|---|---|
| QLoRA (4-bit base) | 15 GB | Yes | 16K tokens | 8 | 256 |
| LoRA (16-bit base) | 34 GB | Yes | 4K tokens | 3 | 128 |
| Full fine-tuning | 240 GB | No | — | — | n/a |
| Inference | Memory at 8K | One L40S? | Longest context | Concurrent 4K sequences |
|---|---|---|---|---|
| 16-bit (BF16/FP16) | 31 GB | Yes | 64K tokens | 20 |
| 8-bit | 16 GB | Yes | 128K tokens | 42 |
| 4-bit | 9.6 GB | Yes | 128K tokens | 51 |
- QLoRA: yes, about 15 GB at 2K tokens, batch 1, rank 16; up to 16K-token sequences at batch 1 or batch 8 at 2K, LoRA rank up to 256.
- LoRA: yes, about 34 GB at 2K tokens, batch 1, rank 16; up to 4K-token sequences at batch 1 or batch 3 at 2K, LoRA rank up to 128.
- Full fine-tuning: no, it needs about 240 GB at 2K tokens, more than the 48 GB available.
Where the memory goes
| Method | Weights | Training state | Activations | Total | On one L40S | 80 GB GPUs needed |
|---|---|---|---|---|---|---|
| QLoRA (4-bit base) | 9.9 GB | 1.0 GB | 3.8 GB | 15 GB | Fits | 1 |
| LoRA (16-bit base) | 30 GB | 1.0 GB | 3.8 GB | 34 GB | Fits | 1 |
| Full fine-tuning | 30 GB | 207 GB | 3.8 GB | 240 GB | Doesn't fit | 4 |
| Precision | Weights | KV cache at 8K | Total | At 128K context | On one L40S |
|---|---|---|---|---|---|
| 16-bit (BF16/FP16) | 30 GB | 1.3 GB | 31 GB | 51 GB | Fits |
| 8-bit | 15 GB | 1.3 GB | 16 GB | 36 GB | Fits |
| 4-bit | 8.3 GB | 1.3 GB | 9.6 GB | 30 GB | Fits |
The cheapest way to run it
Renting an L40S costs $1.09 an hour on Runpod Secure Cloud ($0.79 on Runpod Community Cloud). LoRA on it processes about 2,451 tokens a second by our estimate, so 10 million training tokens take 1.1 hours and cost about $1.24.
| Job (default settings) | Memory | Cheapest rental that fits | Per hour |
|---|---|---|---|
| QLoRA fine-tuning | 15 GB | RTX 3090 | $0.50 |
| LoRA fine-tuning | 34 GB | L40S (this GPU) | $1.09 |
| 4-bit inference | 9.6 GB | RTX 3090 | $0.50 |
| 16-bit inference | 31 GB | L40S (this GPU) | $1.09 |
On-demand rates seen on 15 November 2026: Runpod Secure Cloud, Lambda for the A100 40GB. Consumer cards count as single GPUs only; data-centre GPUs up to eight per node. Kojen's GPU training can launch the L40S on Runpod, Vast.ai or Lambda, or pick the cheapest GPU in stock for you.
Try other settings
| Job | Memory | L40S (48 GB) | Smallest single GPU |
|---|---|---|---|
| QLoRA (4-bit base)2K tokens × batch 1, rank 16 | 15 GB | Fits | RTX 3090 (24 GB) |
| LoRA (16-bit base)2K tokens × batch 1, rank 16 | 34 GB | Fits | A100 40GB (40 GB) |
| Full fine-tuning2K tokens × batch 1 | 240 GB | 6 GPUs | None: needs several |
| Inference, 16-bit (BF16/FP16)8K-token context × 1 sequence | 31 GB | Fits | A100 40GB (40 GB) |
| Inference, 8-bit8K-token context × 1 sequence | 16 GB | Fits | RTX 3090 (24 GB) |
| Inference, 4-bit8K-token context × 1 sequence | 9.6 GB | Fits | RTX 3090 (24 GB) |
A job fits when the estimate is within 90% of the card's memory. Training assumes gradient checkpointing and FlashAttention; multi-GPU counts assume the memory splits evenly (FSDP, ZeRO-3 or tensor parallelism).
How we estimate it
Every figure comes from the model's published architecture (layers, hidden size, attention and KV heads, vocabulary) rather than a rule of thumb, using the same functions as the calculator above. Sizes are in GB (10⁹ bytes).
- Full fine-tuning: 16 bytes per parameter for mixed-precision Adam (bf16 weights and gradients, fp32 master weights and two moments), as in the ZeRO paper.
- LoRA: frozen bf16 weights (2 bytes per parameter) plus 16 bytes per adapter parameter, rank 16 on every linear layer. For mixture-of-experts models, adapters go on attention only and the experts stay frozen.
- QLoRA: NF4 weights with double quantisation (4.127 bits per parameter); embeddings and the output head stay in bf16, as bitsandbytes leaves them.
- Activations: gradient checkpointing and FlashAttention. Each layer's bf16 input is kept, one layer is recomputed at a time, and the output logits take 8 bytes per vocabulary entry per token (fp32 values and gradients). Chunked cross-entropy, as in Unsloth or Liger kernels, cuts that last term.
- Inference: 2, 1 or about 0.56 bytes per weight (4-bit with group-wise scales), plus an fp16 KV cache of 2 × layers × KV heads × head dim × 2 bytes per token; sliding-window layers only cache their window.
- Fits means the estimate is within 90% of the card's listed memory; the rest covers the CUDA context, kernels and fragmentation. Multi-GPU counts assume memory splits evenly.
Speed estimates use L40S's dense BF16 peak of 362 TFLOPS, about 4 FLOPs per active parameter per token for LoRA and 40% utilisation; see fine-tuning and training costs for the method.
Frequently asked questions
- Can I fine-tune Qwen3 14B on an L40S?
- QLoRA: yes, about 15 GB at 2K tokens, batch 1, rank 16; up to 16K-token sequences at batch 1 or batch 8 at 2K, LoRA rank up to 256. LoRA: yes, about 34 GB at 2K tokens, batch 1, rank 16; up to 4K-token sequences at batch 1 or batch 3 at 2K, LoRA rank up to 128. Full fine-tuning: no, it needs about 240 GB at 2K tokens, more than the 48 GB available.
- Can I run Qwen3 14B on an L40S?
- 16-bit (BF16/FP16) inference: yes, 30 GB of weights, room for 64K tokens of context or 20 concurrent 4K-token sequences. 8-bit inference: yes, 15 GB of weights, room for 128K tokens of context or 42 concurrent 4K-token sequences. 4-bit inference: yes, 8.3 GB of weights, room for 128K tokens of context or 51 concurrent 4K-token sequences.
- What does fine-tuning Qwen3 14B on an L40S cost?
- With LoRA at our planning rate of 40% utilisation, about 1.1 hours for 10 million training tokens, around $1.24 at $1.09 an hour (Runpod Secure Cloud, November 2026).
Related
- Qwen3 14B VRAM requirementsEvery GPU and method
- Qwen3 14B on RTX 309024 GB
- Qwen3 14B on RTX 409024 GB
- Qwen3 14B on RTX 509032 GB
- Qwen3 14B on H100 SXM80 GB
- Fine-tuning Qwen3 14B: costPer million tokens
- Llama 3.1 8B on L40SMeta
- Qwen3 8B on L40SAlibaba Qwen
- Qwen3 32B on L40SAlibaba Qwen
- Qwen3 30B-A3B on L40SAlibaba Qwen
- Gemma 3 27B on L40SGoogle DeepMind
Sources
- Hugging Face. Qwen/Qwen3-14B config.json
- Alibaba Qwen. Qwen/Qwen3-14B model card
- Alibaba Qwen. Qwen3 Technical Report (arXiv:2505.09388)
- NVIDIA. L40S specifications
- Runpod. GPU pricing (updated 12 November 2026)
- Lambda. GPU cloud pricing
- Dettmers et al., 2023. QLoRA: Efficient Finetuning of Quantized LLMs (arXiv:2305.14314)
- Hu et al., 2021. LoRA: Low-Rank Adaptation of Large Language Models (arXiv:2106.09685)
- Rajbhandari et al., 2019. ZeRO: Memory Optimizations Toward Training Trillion Parameter Models (arXiv:1910.02054)
- Korthikanti et al., 2022. Reducing Activation Recomputation in Large Transformer Models (arXiv:2205.05198)
Measure before you rent the GPU.
Prompt tuning and model search in Kojen run on hosted models against your own test cases, so you know whether fine-tuning Qwen3 14B is worth the GPU time.
Start training free