Vision Analysis
All Hardware

NVIDIA RTX 5070 Ti

NVIDIA GeForce RTX 5070 Ti

VRAM

16 GB

FP32 Performance

44 TFLOPS

Power (TDP)

300W

Runtime:
Highest Accuracy
Top 5 models by mAP@50-95
61.4%
61.3%
61.1%
60.1%
60.0%
Fastest
Top 5 models by throughput
62 FPS
54 FPS
51 FPS
50 FPS
50 FPS
All Models
55 models benchmarked on NVIDIA RTX 5070 Ti (PyTorch FP32)
#ModelmAPFPSLatencyParams
1
dfine-x
61.4%18.155.1ms62.6M
2
deimv2-x
61.3%17.457.6ms51.2M
3
ec-x
61.1%23.043.5ms49.9M
4
ec-l
60.1%24.241.3ms33.0M
5
rtdetrv4-x
60.0%16.660.1ms62.6M
6
dfine-l
60.0%21.945.8ms31.2M
7
deim-x
59.6%15.066.7ms62.6M
8
deimv2-l
58.6%19.750.7ms32.5M
9
rfdetr-l
58.5%25.139.9ms33.9M
10
ec-m
58.4%26.537.7ms19.4M
11
rtdetr-x
57.9%21.846.0ms67.4M
12
deim-l
57.8%18.653.8ms31.2M
13
dfine-m
57.8%28.834.8ms19.6M
14
rtdetrv4-l
57.8%24.740.4ms31.2M
15
rfdetr-m
57.3%30.333.0ms33.7M
16
yolov9c
57.1%42.323.6ms25.5M
17
rtdetr-r101
56.8%23.642.4ms76.6M
18
rtdetrv2-r101
56.8%25.739.0ms76.6M
19
rtdetrv4-m
56.5%33.629.8ms19.6M
20
yolox-x
56.3%39.625.2ms99.1M
21
yolov9m
56.1%37.126.9ms20.1M
22
deimv2-m
56.0%22.145.4ms18.4M
23
rtdetr-r50
55.9%29.833.6ms42.9M
24
rtdetr-l
55.8%27.736.1ms32.9M
25
rtdetrv2-r50
55.7%32.131.1ms42.9M
26
deim-m
55.5%28.435.3ms19.6M
27
yolox-l
55.4%42.823.3ms54.2M
28
rfdetr-s
55.1%35.827.9ms32.1M
29
rtdetrv2-r50m
54.8%39.925.1ms36.6M
30
ec-s
54.4%27.536.4ms9.9M
31
rtdetr-r50m
53.8%35.328.4ms36.6M
32
dfine-s
53.4%34.229.3ms10.3M
33
rtdetrv2-r34
53.2%40.424.8ms31.4M
34
deimv2-s
53.0%24.341.2ms9.8M
35
rtdetrv4-s
52.9%39.825.1ms10.3M
36
rtdetr-r34
52.2%40.224.9ms31.4M
37
deim-s
52.1%33.430.0ms10.3M
38
yolox-m
51.7%48.220.7ms25.3M
39
rfdetr-n
51.4%39.525.3ms30.5M
40
rtdetrv2-r18
50.8%51.119.6ms20.2M
41
yolov9s
50.5%30.832.5ms7.2M
42
rtdetr-r18
49.8%47.221.2ms20.2M
43
deim-n
46.8%37.027.0ms3.8M
44
deimv2-n
46.7%35.128.5ms3.6M
45
dfine-n
45.8%32.530.7ms3.8M
46
yolox-s
44.3%50.020.0ms9.0M
47
picodet-l
44.1%39.525.3ms3.3M
48
deimv2-pico
42.3%40.524.7ms1.5M
49
yolov9t
41.8%31.931.3ms2.0M
50
picodet-m
37.9%45.122.2ms2.1M
51
yolox-tiny
35.5%61.616.2ms5.1M
52
deimv2-femto
34.5%45.322.1ms1.0M
53
picodet-s
30.4%50.020.0ms1.0M
54
yolox-nano
28.8%53.618.7ms0.9M
55
deimv2-atto
27.5%40.424.7ms0.5M

Run any model with one line

LibreYOLO has the best catalogue of state-of-the-art detectors, all behind one MIT-licensed Python API.

from libreyolo import LibreYOLO, SAMPLE_IMAGE

# LibreYOLO has the best catalogue of state-of-the-art models.
model = LibreYOLO("LibreRFDETRl.pt")           # RF-DETR-L (transformer flagship)
results = model(SAMPLE_IMAGE, save=True)        # run inference, save the annotated image

# Swap in any other model, same one-line API (weights auto-download):
#   LibreYOLO("LibreYOLO9c.pt")      # YOLO9-C
#   LibreYOLO("LibreYOLOXx.pt")      # YOLOX-X
#   LibreYOLO("LibreDFINEx.pt")      # D-FINE-X
#   LibreYOLO("LibreRTDETRr50.pt")   # RT-DETR-R50