NVIDIA · Ampere · unified memory

What runs on a Jetson AGX Orin 64GB

64 GB at 205 GB/s. The largest model that fits at Q4_K_M is Qwen2.5 72B, at roughly 3.02 tokens per second. Unified memory means the operating system takes a share, so budget around three quarters of the total.

Runs comfortably

6.1 GB of 48.0 GB · 13%
048 GB
Weights 4.6 GB
KV cache 1.0 GB
Runtime overhead 0.5 GB

Llama 3.1 8B at Q4_K_M leaves 41.9 GB spare on a Jetson AGX Orin 64GB. There is room to raise the context length or move up a quantisation level.

Generation25.0tok/s
Prompt processing1098tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Llama 3.1 8B on a Jetson AGX Orin 64GB

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 15.0 GB 16.5 GB Runs comfortably 128K 8.29
INT8 / W8A8 8.50 7.7 GB 9.2 GB Runs comfortably 128K 15.6
Q8_0 (GGUF) 8.50 7.7 GB 9.2 GB Runs comfortably 128K 15.6
FP8 (E4M3) 8.00 7.5 GB 9.0 GB Runs comfortably 128K 16.0
Q6_K 6.56 6.1 GB 7.7 GB Runs comfortably 128K 19.3
AWQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 21.6
GPTQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 21.6
MXFP4 4.25 5.4 GB 6.9 GB Runs comfortably 128K 21.6
Q5_K_M 5.67 5.3 GB 6.8 GB Runs comfortably 128K 22.0
Q5_K_S 5.52 5.2 GB 6.7 GB Runs comfortably 128K 22.6
Q4_K_M 4.85 4.6 GB 6.1 GB Runs comfortably 128K 25.0
Q4_K_S 4.58 4.4 GB 5.9 GB Runs comfortably 128K 26.1
Q4_0 4.55 4.4 GB 5.9 GB Runs comfortably 128K 26.2
IQ4_XS 4.25 4.1 GB 5.6 GB Runs comfortably 128K 27.6
Q3_K_M 3.91 3.8 GB 5.4 GB Runs comfortably 128K 29.3
IQ3_M 3.70 3.7 GB 5.2 GB Runs comfortably 128K 30.5
IQ3_XXS 3.06 3.2 GB 4.7 GB Runs comfortably 128K 34.8
Q2_K 2.63 2.8 GB 4.3 GB Runs comfortably 128K 38.5
IQ2_XXS 2.06 2.3 GB 3.9 GB Runs comfortably 128K 44.7
IQ1_M 1.75 2.1 GB 3.6 GB Runs comfortably 128K 49.0

Models on a Jetson AGX Orin 64GB

Q4_K_M at 8K context. Bandwidth sets the speed; capacity sets the ceiling.

ModelParamsWeightsVerdictMax ctxtok/s
DeepSeek-R1 671B-A37B 671B 379.0 GB Won't fit — 2.02
Qwen3 235B-A22B 235.1B 132.8 GB Won't fit — 3.74
gpt-oss 120B-A5.1B 116.8B 66.0 GB Won't fit — 21.9
Llama 4 Scout 109B-A17B 109B 61.7 GB Won't fit — 7.18
Qwen2.5 72B 72.7B 41.2 GB Fits, but tight 20K 3.02
Llama 3.1 70B 70.5B 40.0 GB Runs comfortably 24K 3.11
Mixtral 8x7B 46.7B 26.4 GB Runs comfortably 32K 14.0
Command R 35B 35.0B 20.1 GB Runs comfortably 22K 5.11
Yi-1.5 34B 34.4B 19.5 GB Runs comfortably 32K 6.27
Qwen3 32B 32.8B 18.6 GB Runs comfortably 115K 6.53
Qwen2.5-Coder 32B 32.8B 18.6 GB Runs comfortably 115K 6.53
DeepSeek-R1-Distill-Qwen 32B 32.8B 18.6 GB Runs comfortably 115K 6.53
Qwen3 30B-A3B 30.5B 17.3 GB Runs comfortably 128K 40.6
Gemma 3 27B 27.4B 15.6 GB Runs comfortably 128K 7.84
Mistral Small 24B 23.6B 13.4 GB Runs comfortably 32K 9.12
gpt-oss 20B-A3.6B 20.9B 11.9 GB Runs comfortably 128K 47.0
Qwen3 14B 14.8B 8.5 GB Runs comfortably 128K 14.1
Phi-4 14B 14.7B 8.4 GB Runs comfortably 16K 14.0
Mistral NeMo 12B 12.3B 7.0 GB Runs comfortably 128K 16.7
Gemma 3 12B 12.2B 7.0 GB Runs comfortably 128K 17.0
GLM-4 9B 9.4B 5.4 GB Runs comfortably 128K 23.0
Qwen3 8B 8.2B 4.7 GB Runs comfortably 128K 24.2
Llama 3.1 8B 8.0B 4.6 GB Runs comfortably 128K 25.0
DeepSeek-R1-Distill-Llama 8B 8.0B 4.6 GB Runs comfortably 128K 25.0
Qwen2.5 7B 7.6B 4.4 GB Runs comfortably 128K 27.7
Mistral 7B v0.3 7.3B 4.1 GB Runs comfortably 32K 27.7
Gemma 3 4B 4.3B 2.5 GB Runs comfortably 128K 47.4
Qwen3 4B 4.0B 2.3 GB Runs comfortably 128K 44.4
Llama 3.2 3B 3.2B 1.8 GB Runs comfortably 128K 55.8
Llama 3.2 1B 1.2B 0.7 GB Runs comfortably 128K 150

Direct answers

Llama 3.1 8B on Jetson AGX Orin 64GB
8.0BLlama 3.1 70B on Jetson AGX Orin 64GB
70.5BLlama 3.2 3B on Jetson AGX Orin 64GB
3.2BLlama 3.2 1B on Jetson AGX Orin 64GB
1.2BLlama 4 Scout 109B-A17B on Jetson AGX Orin 64GB
109BQwen3 8B on Jetson AGX Orin 64GB
8.2BQwen3 14B on Jetson AGX Orin 64GB
14.8BQwen3 32B on Jetson AGX Orin 64GB
32.8B