Apple M4 (10-core GPU, 120 GB/s) vs NVIDIA Jetson AGX Orin 64GB

Apple Mac mini (M4, 2024), small silver aluminium desktop, front view — Apple Mac mini (M4, 2024) — the M4 Pro model shares this enclosure. Photo: Seasider53, CC BY 4.0, via Wikimedia Commons.

NVIDIA Jetson AGX Orin developer kit with its retail box, on a black background — NVIDIA Jetson AGX Orin developer kit. Photo: Auledas, CC BY 4.0, via Wikimedia Commons.

Showing the newest results per platform and source. 12 superseded older results are hidden.Showing all results, including 12 superseded by newer results from the same source.

NVIDIA Jetson AGX Orin 64GB is faster in all 3 like-for-like comparisons (largest gap: LLM generation, 1.45×).

	Apple M4 (10-core GPU, 120 GB/s)	NVIDIA Jetson AGX Orin 64GB
Best object detection	357 FPS B 2026-10-01 (page last edited)	435 FPS B 2026-10-13 (page last edited)
Power	n/a	65.00 W spec
Perf / W	n/a	6.69 FPS/W spec
Price	n/a	$3499 list
Workloads with data	6 of 7	3 of 7

Best published object detection result per platform (may be from different like-for-like groups — see data). n/a = not published; never estimated.

Show data — all workloads, sources and like-for-like groups

object detection

✓ Like-for-like: model-size class nano (YOLO-n), batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	435 FPS inference_time · b=1	YOLO26n INT8 · Ultralytics TensorRT · nano (YOLO-n)	B	65.00 W spec platform spec envelope (not measured with this result)	Ultralytics documentation (benchmark table) 2026-10-13 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	357 FPS latency_e2e · b=1	YOLO26n FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

instance segmentation

Only Apple M4 (10-core GPU, 120 GB/s) has published instance segmentation results.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	196 FPS latency_e2e · b=1	YOLO26n-seg FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

NVIDIA Jetson AGX Orin 64GB:

◌

No published instance segmentation results for NVIDIA Jetson AGX Orin 64GB or Apple M4 (10-core GPU, 120 GB/s) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

pose estimation

Only Apple M4 (10-core GPU, 120 GB/s) has published pose estimation results.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	370 FPS latency_e2e · b=1	YOLO26n-pose FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

NVIDIA Jetson AGX Orin 64GB:

◌

No published pose estimation results for NVIDIA Jetson AGX Orin 64GB or Apple M4 (10-core GPU, 120 GB/s) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

image classification

Only Apple M4 (10-core GPU, 120 GB/s) has published image classification results.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	417 FPS latency_e2e · b=1	YOLO26n-cls FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

NVIDIA Jetson AGX Orin 64GB:

◌

No published image classification results for NVIDIA Jetson AGX Orin 64GB or Apple M4 (10-core GPU, 120 GB/s) yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

semantic segmentation

⚠ MIXED COMPARISON — not like-for-like. Baseline nano (YOLO-n) at batch 1. Differs: model-size class unknown vs nano (YOLO-n).

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	53.48 FPS latency_e2e · b=1	YOLO26n-sem FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)
NVIDIA Jetson AGX Orin 64GB MIXED: model-size class unknown vs nano (YOLO-n)	104 FPS throughput · b=1	SegFormer-B0 (640x640) FP16 · TensorRT (ONNX export) · unknown	C	65.00 W spec platform spec envelope (not measured with this result)	arXiv 2608.00927 AgriJetsonBench (abstract figures) 2026-08-02 (published)

No shared like-for-like group exists for this pair.

speech-to-text

◌

No published speech-to-text results for Apple M4 (10-core GPU, 120 GB/s) or NVIDIA Jetson AGX Orin 64GB yet.
We only show cited public figures — nothing is estimated to fill gaps. Requests tell us what to source or test next.

LLM generation

✓ Like-for-like: model-size class LLM 5–14B, batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	30.00 tok/s generation_throughput · b=1	Qwen 3 8B W4A16 · vLLM · LLM 5–14B	C	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson AI Lab benchmarks (benchmarks.json) 2026-07-21 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	24.11 tok/s tg128 · b=1	Llama 2 7B Q4_0 · llama.cpp Metal (commit 8e672ef) · LLM 5–14B	C	unknown power n/a no power published	llama.cpp discussion #4167 scoreboard (via GitHub API) 2026-10-01 (page last edited)

✓ Like-for-like: model-size class LLM ≤4B, batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	60.00 tok/s generation_throughput · b=1	Qwen3-VL-2B W4A16 · vLLM · LLM ≤4B	C	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson AI Lab benchmarks (benchmarks.json) 2026-07-21 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	41.31 tok/s generation_throughput · b=1	Llama 3.2:3b as in linked issue (Ollama default tags are 4-bit) · Ollama / llama.cpp (see linked issue) · LLM ≤4B	C	30.10 W measured reported in result source — measured	Jeff Geerling ai-benchmarks README (Findings tables) 2026-09-22 (page last edited)

NVIDIA Jetson AGX Orin 64GB is faster in all 3 like-for-like comparisons (largest gap: LLM generation, 1.45×).

	Apple M4 (10-core GPU, 120 GB/s)	NVIDIA Jetson AGX Orin 64GB
Best object detection	357 FPS B 2026-10-01 (page last edited)	435 FPS B 2026-10-13 (page last edited)
Power	n/a	65.00 W spec
Perf / W	n/a	6.69 FPS/W spec
Price	n/a	$3499 list
Workloads with data	6 of 7	5 of 7

Best published object detection result per platform (may be from different like-for-like groups — see data). n/a = not published; never estimated.

Show data — all workloads, sources and like-for-like groups

object detection

✓ Like-for-like: model-size class nano (YOLO-n), batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	435 FPS inference_time · b=1	YOLO26n INT8 · Ultralytics TensorRT · nano (YOLO-n)	B	65.00 W spec platform spec envelope (not measured with this result)	Ultralytics documentation (benchmark table) 2026-10-13 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	357 FPS latency_e2e · b=1	YOLO26n FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

instance segmentation

Only Apple M4 (10-core GPU, 120 GB/s) has published instance segmentation results.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	196 FPS latency_e2e · b=1	YOLO26n-seg FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

NVIDIA Jetson AGX Orin 64GB:

◌

pose estimation

Only Apple M4 (10-core GPU, 120 GB/s) has published pose estimation results.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	370 FPS latency_e2e · b=1	YOLO26n-pose FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)

NVIDIA Jetson AGX Orin 64GB:

◌

image classification

⚠ MIXED COMPARISON — not like-for-like. Baseline nano (YOLO-n) at batch 1. Differs: model-size class mid-size CNN vs nano (YOLO-n); batch unknown vs 1.

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	417 FPS latency_e2e · b=1	YOLO26n-cls FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)
NVIDIA Jetson AGX Orin 64GB MIXED: model-size class mid-size CNN vs nano (YOLO-n); batch unknown vs 1	6424 FPS throughput_offline · b=unknown	MLPerf ResNet INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · mid-size CNN	D	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) Older data (2023)

No shared like-for-like group exists for this pair.

semantic segmentation

⚠ MIXED COMPARISON — not like-for-like. Baseline nano (YOLO-n) at batch 1. Differs: model-size class unknown vs nano (YOLO-n).

Platform	Result	Model · precision · stack	Grade	Power	Source
Apple M4 (10-core GPU, 120 GB/s)	53.48 FPS latency_e2e · b=1	YOLO26n-sem FP16 · Ultralytics 8.4.168 Core AI (coreai-torch 0.4.3), macOS 27.0 · nano (YOLO-n)	B	unknown power n/a no power published	Ultralytics documentation: Apple Core AI integration benchmark 2026-10-01 (page last edited)
NVIDIA Jetson AGX Orin 64GB MIXED: model-size class unknown vs nano (YOLO-n)	104 FPS throughput · b=1	SegFormer-B0 (640x640) FP16 · TensorRT (ONNX export) · unknown	C	65.00 W spec platform spec envelope (not measured with this result)	arXiv 2608.00927 AgriJetsonBench (abstract figures) 2026-08-02 (published)

No shared like-for-like group exists for this pair.

speech-to-text

Only NVIDIA Jetson AGX Orin 64GB has published speech-to-text results.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	1170 FPS throughput_offline · b=unknown	MLPerf RNN-T INT8/FP16 (MLPerf closed) · TensorRT (MLPerf Inference, older round) · RNN-T	D	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson Benchmarks (MLPerf v3.1 results table) Older data (2023)

Apple M4 (10-core GPU, 120 GB/s):

◌

LLM generation

✓ Like-for-like: model-size class LLM 5–14B, batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	30.00 tok/s generation_throughput · b=1	Qwen 3 8B W4A16 · vLLM · LLM 5–14B	C	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson AI Lab benchmarks (benchmarks.json) 2026-07-21 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	24.11 tok/s tg128 · b=1	Llama 2 7B Q4_0 · llama.cpp Metal (commit 8e672ef) · LLM 5–14B	C	unknown power n/a no power published	llama.cpp discussion #4167 scoreboard (via GitHub API) 2026-10-01 (page last edited)

✓ Like-for-like: model-size class LLM ≤4B, batch 1. Model, precision and stack may still differ.

Platform	Result	Model · precision · stack	Grade	Power	Source
NVIDIA Jetson AGX Orin 64GB	60.00 tok/s generation_throughput · b=1	Qwen3-VL-2B W4A16 · vLLM · LLM ≤4B	C	65.00 W spec platform spec envelope (not measured with this result)	NVIDIA Jetson AI Lab benchmarks (benchmarks.json) 2026-07-21 (page last edited)
Apple M4 (10-core GPU, 120 GB/s)	41.31 tok/s generation_throughput · b=1	Llama 3.2:3b as in linked issue (Ollama default tags are 4-bit) · Ollama / llama.cpp (see linked issue) · LLM ≤4B	C	30.10 W measured reported in result source — measured	Jeff Geerling ai-benchmarks README (Findings tables) 2026-09-22 (page last edited)

How to read these numbers

Every number is labelled. Confidence: A MLPerf closed, audited · B documented third-party method · C community / review · D vendor-published, not independently verified · E estimate. Power basis: measured with the result · spec vendor spec/envelope · power n/a not published.

Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.