Model Library/Llama 3.3 70B

Meta

Llama 3.3 70B

A dense 70B model. Slow for one user on a Spark, useful for batch work.

Textllama.cpp2 builds

Recommended: DGX Spark, 128 GB unified, $0.79 / hour, billed per minute. Check availability.

Configure

Every build below is published on Hugging Face and fits at least one GPU we rent. Pick the quantization and engine: the memory it needs, the GPUs it fits and the recommended GPU follow, and Deploy takes that exact build, revision and GPU to the console. Only the one-click preset is tested by GPUwerk.

Llama 3.3 70B, Q4_K_M GGUF on llama.cpp
QuantizationQ4_K_M GGUF
Enginellama.cpp
StatusUntested by GPUwerk
Weights42.5 GB
GPU memory it needsAbout 51.9 GB with a 32k-token cache, an estimate
Recommended GPUDGX Spark, $0.79 / hour
Weights sourceunsloth/Llama-3.3-70B-Instruct-GGUF, revision 8f14c5c5d06f, file Llama-3.3-70B-Instruct-Q4_K_M.gguf
Licence on the cardllama3.3
GPUs Llama 3.3 70B Q4_K_M GGUF does and does not fit
GPUMemoryPriceAvailabilityDeploy
RTX 5090x86 · Blackwell32 GB GDDR7Does not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.
DGX SparkArm · GB10 · recommended128 GB unified$0.79 / hourCheck availabilityRun it yourself Llama 3.3 70B Q4_K_M GGUF on DGX Spark
RTX PRO 6000x86 · Blackwell96 GB GDDR7Does not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.
H100 SXMx86 · Hopper80 GB HBM3Does not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.
H200x86 · Hopper141 GB HBM3eDoes not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.
B200x86 · Blackwell180 GB HBM3eDoes not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.
B300x86 · Blackwell288 GB HBM3eDoes not fit: GGUF models run on llama.cpp, which has no x86_64 build yet; pick a DGX Spark, or a safetensors build for vLLM.

Builds by the model's maker, NVIDIA and Unsloth, read from Hugging Face on 21 November 2026. Memory is an estimate (weights x 1.15 plus 3 GB); GGUF runs on llama.cpp, which GPUwerk builds for the DGX Spark only, and NVFP4 needs a Blackwell GPU. SGLang is not offered: there is no SGLang runtime yet.

Details

About the model.

Llama 3.3 70B
MakerMeta
TaskText
Parameters70B, dense
LicenceSee the model card

6.0 tok/s for one request on one DGX Spark, measured by GPUwerk with vLLM on 2 September 2026. Every other speed figure is in the model guide, with its source.

Quick start

  1. Deploy a DGX Spark with a base image (DGX OS or Ubuntu).
  2. Follow the Llama 3.3 70B guide to download the weights and start vLLM.
  3. Connect over SSH or the browser terminal; the machine is yours until you terminate it.

SSH, the browser terminal, VNC and your workspace work the same on every instance. Terminate when you are done and billing ends.

Related models

QwenCoding · Text

Qwen3 Coder 30B

A coding model for agents and editors, with tool calls turned on.

One-click preset · vLLM · RTX 5090

QwenText

Qwen3.6 35B-A3B

A fast mixture-of-experts model for chat and agents, also on the pay-per-token API.

One-click preset · llama.cpp · DGX Spark

OpenAIText

GPT-OSS 120B

OpenAI's open-weight 120B model, quantized to fit one DGX Spark.

One-click preset · llama.cpp · DGX Spark

All models in the library