Apertus 1.5 70B — the Swiss AI Initiative's fully open dense model, with published data and training code. Version 1.5 adds image and audio input, an optional thinking mode, tool use and a 262K context.
chatmultilingualvisionaudioreasoningtool_use
72.01B
Parameters
256K
Context length
0
Benchmarks
16
Quantizations
8K
HF downloads
Architecture
Dense
Released
2026-07-24
Layers
80
KV Heads
8
Head Dim
128
Family
other
Quantization Options
Context length:
Quant
Bits
VRAM @ 16K
Quality
IQ2_M
2.93
30.6 GB
26.9 + 3.8 KV
low
Q2_K
3.16
32.7 GB
28.9 + 3.8 KV
low
IQ3_XXS
3.25
33.5 GB
29.7 + 3.8 KV
low
IQ3_XS
3.5
35.7 GB
32.0 + 3.8 KV
low
Q3_K_S
3.64
37.0 GB
33.3 + 3.8 KV
low
IQ3_M
3.76
38.1 GB
34.3 + 3.8 KV
low
Q3_K_M
4
40.2 GB
36.5 + 3.8 KV
low
Q3_K_L
4.3
42.9 GB
39.2 + 3.8 KV
moderate
IQ4_XS
4.46
44.4 GB
40.6 + 3.8 KV
moderate
Q4_K_S
4.67
46.3 GB
42.5 + 3.8 KV
moderate
Q4_K_M
4.89
48.3 GB
44.5 + 3.8 KV
good
Q5_K_S
5.57
54.4 GB
50.6 + 3.8 KV
good
Q5_K_M
5.7
55.5 GB
51.8 + 3.8 KV
good
Q6_K
6.56
63.3 GB
59.5 + 3.8 KV
excellent
Q8_0
8.5
80.7 GB
77.0 + 3.8 KV
lossless
FP16
16
148.3 GB
144.5 + 3.8 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.