Skip to content

VRAM reference

Llama 3.3 70B on an H100 SXM

Yes, you can fine-tune Llama 3.3 70B on one H100 SXM (80 GB) with QLoRA. 4-bit inference: yes, 40 GB of weights, room for 64K tokens of context or 24 concurrent 4K-token sequences.

By Will Schulz · Checked 15 November 2026 · Prices as of November 2026

Verdict by method

Fine-tuningMemory at 2K tokensOne H100 SXM?Longest sequence (batch 1)Largest batch at 2KLargest LoRA rank
QLoRA (4-bit base)48 GBYes8K tokens5128
LoRA (16-bit base)150 GBNo———
Full fine-tuning1,134 GBNo——n/a
InferenceMemory at 8KOne H100 SXM?Longest contextConcurrent 4K sequences
16-bit (BF16/FP16)144 GBNo——
8-bit73 GBShort contexts only4K tokens1
4-bit42 GBYes64K tokens24
  • QLoRA: yes, about 48 GB at 2K tokens, batch 1, rank 16; up to 8K-token sequences at batch 1 or batch 5 at 2K, LoRA rank up to 128.
  • LoRA: no, it needs about 150 GB at 2K tokens, more than the 80 GB available.
  • Full fine-tuning: no, it needs about 1,134 GB at 2K tokens, more than the 80 GB available.

Where the memory goes

At 2,048 tokens per sequence, batch 1, LoRA rank 16, gradient checkpointing on
MethodWeightsTraining stateActivationsTotalOn one H100 SXM80 GB GPUs needed
QLoRA (4-bit base)40 GB3.3 GB5.6 GB48 GBFits1
LoRA (16-bit base)141 GB3.3 GB5.6 GB150 GBDoesn't fit3
Full fine-tuning141 GB988 GB5.6 GB1,134 GBDoesn't fit16
One sequence, fp16 KV cache
PrecisionWeightsKV cache at 8KTotalAt 128K contextOn one H100 SXM
16-bit (BF16/FP16)141 GB2.7 GB144 GB184 GBDoesn't fit
8-bit71 GB2.7 GB73 GB114 GBDoesn't fit
4-bit40 GB2.7 GB42 GB83 GBFits

The cheapest way to run it

Renting an H100 SXM costs $4.39 an hour on Runpod Secure Cloud ($2.69 on Runpod Community Cloud). QLoRA on it processes about 1,093 tokens a second by our estimate, so 10 million training tokens take 2.5 hours and cost about $11.15.

Job (default settings)MemoryCheapest rental that fitsPer hour
QLoRA fine-tuning48 GBA100 80GB$2.19
LoRA fine-tuning150 GB4 × L40S$4.36
4-bit inference42 GBL40S$1.09
16-bit inference144 GB2 × A100 80GB$4.38

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 H100 SXM on Runpod, Vast.ai or Lambda, or pick the cheapest GPU in stock for you.

Try other settings

JobMemoryH100 SXM (80 GB)Smallest single GPU
QLoRA (4-bit base)2K tokens × batch 1, rank 1648 GBFitsA100 80GB (80 GB)
LoRA (16-bit base)2K tokens × batch 1, rank 16150 GB3 GPUsB200 (180 GB)
Full fine-tuning2K tokens × batch 11,134 GB16 GPUsNone: needs several
Inference, 16-bit (BF16/FP16)8K-token context × 1 sequence144 GB2 GPUsB200 (180 GB)
Inference, 8-bit8K-token context × 1 sequence73 GB2 GPUsH200 (141 GB)
Inference, 4-bit8K-token context × 1 sequence42 GBFitsL40S (48 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 H100 SXM's dense BF16 peak of 989 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 Llama 3.3 70B on an H100 SXM?
QLoRA: yes, about 48 GB at 2K tokens, batch 1, rank 16; up to 8K-token sequences at batch 1 or batch 5 at 2K, LoRA rank up to 128. LoRA: no, it needs about 150 GB at 2K tokens, more than the 80 GB available. Full fine-tuning: no, it needs about 1,134 GB at 2K tokens, more than the 80 GB available.
Can I run Llama 3.3 70B on an H100 SXM?
16-bit (BF16/FP16) inference: no, the weights alone are 141 GB. 8-bit inference: yes, 71 GB of weights, room for 4K tokens of context. 4-bit inference: yes, 40 GB of weights, room for 64K tokens of context or 24 concurrent 4K-token sequences.
What does fine-tuning Llama 3.3 70B on an H100 SXM cost?
With QLoRA at our planning rate of 40% utilisation, about 2.5 hours for 10 million training tokens, around $11.15 at $4.39 an hour (Runpod Secure Cloud, November 2026).

Sources

  1. [1]Hugging Face. meta-llama/Llama-3.3-70B-Instruct config.json
  2. [2]Meta. meta-llama/Llama-3.3-70B-Instruct model card
  3. [3]Meta. The Llama 3 Herd of Models (arXiv:2407.21783)
  4. [4]NVIDIA. H100 SXM specifications
  5. [5]Runpod. GPU pricing (updated 12 November 2026)
  6. [6]Lambda. GPU cloud pricing
  7. [7]Dettmers et al., 2023. QLoRA: Efficient Finetuning of Quantized LLMs (arXiv:2305.14314)
  8. [8]Hu et al., 2021. LoRA: Low-Rank Adaptation of Large Language Models (arXiv:2106.09685)
  9. [9]Rajbhandari et al., 2019. ZeRO: Memory Optimizations Toward Training Trillion Parameter Models (arXiv:1910.02054)
  10. [10]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 Llama 3.3 70B is worth the GPU time.

Start training free