Cloud Souverain

Flavors

A cloud server instance consists of one or more virtual processors (vCPU), memory, an OS image (operating system, public or private image) and instance block storage (IBS) or volume block storage (Volume Service VBS).

So that the components do not have to be put together individually, you can choose from the preconfigured instance types, called flavors. These flavors combine firmly defined sizes and limits in terms of network bandwidth, storage throughput or IOPS number and type of vCPUs, memory size and instance block storage.

Note

For flavors (e* and m*) that have an instance block storage disk (IBS-Nova), the prices for a root disk already are included in the flavors and are charged depending on the VM power status (UP). The same applies to the p* Flavor with a local SSD disk. If a customer wants to use an e*, m* or p* flavor, but decides to use a volume (cinder) as a root disk when creating a VM, the costs for the volume are added.

In addition to the performance-related features described above, the flavor groups differ fundamentally in terms of whether they include an instance block storage (IBS) (all m* and e* flavors) or whether they use local SSD storage (p* flavors) or whether a cinder volume block storage (volume service VBS) is used as a root disk volume (all zero-disk flavors g*). With the zero-disk flavors (g*), the size and IOPS performance of the volume block storage (VBS) can be selected individually according to your requirements.

General Workload e1

The Flavor Group e1 (Enterprise Multi Purpose Instances) offers a balanced ratio of all components, such as vCPU, RAM, IOPS and network, and thus creates a balance between computing, memory and network resources. It includes a defined instance disk. The ratio of vCPU to RAM is 1:4, making the Flavor Group e1 a good choice for many applications and tasks.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)

  • Up to 24 vCPU's

  • Up to 192GB RAM

  • Up to 100,000 kilobyte/s network bandwidth

  • Instance block storage 10 GB to 120 GB

  • Storage throughput up to 300 MiB/s

  • Instance block storage IOPS up to 2,000

  • Distributed NVMe based storage solution with improved level of IOPS

  • Improved network burst performance for backend servers for EA

Use Cases

  • Development server

  • Web or business applications

  • Small to medium-sized databases

  • Microsoft SharePoint, cluster computing and other business applications

  • Data processing tasks with moderate requirements for network bandwidth, storage size and throughput

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

e1.small

1

2

50

0.07

55.64

38.95

27.82

e1.micro

1

1

20

0.06

47.51

33.26

23.75

e1.nano

1

0.5

10

0.06

44.08

30.86

22.04

e1.medium

2

4

80

0.12

89.45

62.62

44.73

e1.large

2

8

120

0.15

111.71

78.20

55.85

e1.xlarge

4

16

120

0.25

188.75

132.12

94.37

e1.2xlarge

8

32

120

0.46

342.83

239.98

171.41

e1.4xlarge

16

64

120

0.87

650.99

455.69

325.49

e1.6xlarge

24

96

120

1.29

959.15

671.40

479.57

Memory intense Workload e1m

The e1m (Enterprise Memory Intensive Instances) flavor group is particularly suitable for applications that require a lot of memory. It contains a defined instance disk. The ratio of vCPU to RAM is 1:8. Databases and Java applications in particular benefit from additional RAM.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)
  • Up to 24 vCPU's
  • Up to 192GB RAM
  • Up to 100,000 kilobyte/s network bandwidth
  • Instance block storage 10 GB to 120 GB
  • Storage throughput up to 300 MiB/s
  • Instance block storage IOPS up to 2,000
  • Distributed NVMe based storage solution with improved level of IOPS
  • Improved network burst performance for backend servers for EA

Use Cases

  • Big Data
  • In-Memory Caches
  • Medium in-memory databases
  • High performance databases
  • Data processing tasks with higher memory requirements

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

e1m.micro

1

2

20

0.07

51.79

36.25

25.89

e1m.nano

1

1

10

0.06

46.22

32.36

23.11

e1m.small

1

4

50

0.09

64.20

44.94

32.10

e1m.large

2

16

120

0.20

145.95

102.16

72.97

e1m.medium

2

8

80

0.14

106.57

74.60

53.29

e1m.xlarge

4

32

120

0.35

257.23

180.06

128.61

e1m.2xlarge

8

64

120

0.64

479.79

335.85

239.89

e1m.4xlarge

16

128

120

1.24

924.91

647.44

462.45

e1m.6xlarge

24

192

120

1.84

1'370.03

959.02

685.01

CPU intense Workload e1c

The e1c (Enterprise CPU Intensive Instances) flavor group is ideal for compute-intensive applications. It contains a defined instance disk. The ratio of vCPU to RAM is 1:2. This flavor is often used for build servers that compile code or for cryptographic calculations.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)
  • Up to 24 vCPU's
  • Up to 48GB RAM
  • Up to 100,000 kilobyte/s network bandwidth
  • Instance block storage 10 GB to 120 GB
  • Storage throughput up to 300 MiB/s
  • Instance block storage IOPS up to 2,000
  • Distributed NVMe based storage solution with improved level of IOPS
  • Improved network burst performance for backend servers for EA

Use Cases

  • HPC
  • Video encoding
  • Deep Learning
  • Powerful web servers
  • Compute-intensive cluster solutions
  • Scientific modeling
  • Data processing tasks with higher CPU requirements

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1 year monthly rate (CHF)

Pre-Commit 3 years monthly rate (CHF)

e1c.micro

1

0.5

20

0.06

45.37

31.76

22.68

e1c.nano

1

0.25

10

0.06

43.01

30.11

21.51

e1c.small

1

1

50

0.07

51.36

35.95

25.68

e1c.large

2

4

120

0.13

94.59

66.21

47.29

e1c.medium

2

2

80

0.11

80.89

56.62

40.45

e1c.xlarge

4

8

120

0.21

154.51

108.16

77.25

e1c.2xlarge

8

16

120

0.37

274.35

192.04

137.17

e1c.4xlarge

16

32

120

0.69

514.03

359.82

257.01

e1c.6xlarge

24

48

120

1.01

753.71

527.60

376.85

Network intense Workload e1n

The e1n (Enterprise Network Intensive Instances) flavor group is suitable for applications that require a particularly high network throughput. Compared to the other e1 variants, it offers more than twice the network throughput. It contains a defined instance disk.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)
  • Up to 24 vCPU's
  • Up to 96GB RAM
  • Instance block storage 10 GB to 120 GB
  • Up to 250'000 kilobyte/s network bandwidth
  • Distributed NVMe based storage solution with improved level of IOPS
  • Improved network burst performance for backend servers for EA

Use Cases

  • Web or business applications
  • Small to medium-sized databases
  • Game servers
  • Caching, cluster computing and other business applications
  • Data processing tasks with moderate requirements for network bandwidth, storage size and throughput

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

e1n.micro

1

1

20

0.08

60.35

42.24

30.17

e1n.small

1

2

50

0.09

68.48

47.94

34.24

e1n.nano

1

0.5

10

0.08

56.92

39.85

28.46

e1n.large

2

8

120

0.17

124.55

87.18

62.27

e1n.medium

2

4

80

0.14

102.29

71.60

51.15

e1n.xlarge

4

16

120

0.27

201.59

141.11

100.79

e1n.2xlarge

8

32

120

0.48

355.67

248.97

177.83

e1n.4xlarge

16

64

120

0.89

663.83

464.68

331.91

e1n.6xlarge

24

96

120

1.31

971.99

680.39

485.99

Pure Power intense workloads p1d

The new p1d flavor group, based on the advanced AMD Ryzen 9 Zen4 processor, offers an impressive base clock speed of up to 5.7 GHz, making it one of the most powerful processors in the prestigious Ryzen 7000 series. These processors are specifically designed for demanding applications and workloads, from professional graphics editing and 3D modeling to complex scientific computing. In addition, the Flavor Group p1d is equipped with high-performance local SSD storage that enables fast data access and transfers to support even the most demanding, data-intensive applications. With the exceptional computing power and efficiency of the AMD Ryzen 9 Zen4 combined with fast local SSD storage, you can get your projects done faster than ever before without compromising on quality or performance.

Features

  • AMD Ryzen 9 Zen4 CPU
  • Up to 16 vCPU's with up to 5.7 GHz
  • Up to 64GB RAM
  • Local NVMe instance block storage up to 960 GB and 250,000 IOPS
  • Up to 250,000 kilobyte/s network bandwidth
  • Improved network performance through burst mode

Use Cases

  • HPC
  • Video encoding
  • Deep Learning
  • Powerful web servers
  • Compute-intensive cluster solutions
  • Scientific modeling
  • Data processing tasks with higher CPU requirements

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

p1d.small

1

2

50

0.131397849

97.76

68.432

48.88

p1d.medium

1

4

80

0.148080645

110.172

77.1204

55.086

p1d.large

2

8

120

0.211596774

157.428

110.1996

78.714

p1d.xlarge

4

16

240

0.345532258

257.076

179.9532

128.538

p1d.2xlarge

8

32

480

0.613403226

456.372

319.4604

228.186

p1d.4xlarge

16

64

960

1.149145161

854.964

598.4748

427.482

Local storage

It is important to note that local SSD storage offers maximum storage performance, but data stored on a local SSD can be lost if the SSD or the compute host fails. We therefore strongly recommend that you make regular backups of your important data to avoid data loss. In addition, an instance cannot be started if the compute host is unavailable, which underlines the need for a robust backup and recovery strategy. Of course, as usual, you can also create volume-based instances with this flavor group that are not affected by the above limitations.

GPU-Powered Intensive Workloads a1g200d

This flavor was designed for demanding AI and data center workloads and features the high-performance NVIDIA H200 Tensor Core GPU with a generous 141 GB of HBM3e VRAM. This makes it ideal for training and inference of large language models, as well as for computationally intensive HPC simulations. For particularly demanding requirements, the configurable GPU instances a1g200d.16xlarge, a1g200d.32xlarge, and a1g200d.64xlarge are available. It combines up to 192 vCPUs, 768 GB of RAM, and 960 GB of NVMe disk space into an environment that requires no compromises, even for the most extensive workloads. These instances are designed for enterprises and research institutions that require maximum computing power for their AI, machine learning, and scientific applications. At its core is an AMD EPYC 9755 processor with an impressive 128 cores and a base clock speed of up to 4.1 GHz, which ensures reliable performance even under sustained full load.

Features

  • AMD EPYC 9755 128-Core Processor
  • NVIDIA H200 NVL GPU with 141 GB vRAM
  • Up to 192 vCPU's with up to 4.1 GHz
  • Up to 768GB RAM
  • Local NVMe instance block storage up to 960 GB and 250,000 IOPS
  • Up to 20 Gbit/s network bandwidth
  • Improved network performance through burst mode

Use Cases

  • LLM Inference with Long Context Windows
  • LLM Training & Fine-Tuning
  • RAG Pipelines & Embedding Workloads
  • Agentic AI
  • Scientific Simulations
  • Medical Imaging
  • Financial Models & Fraud Detection
  • MIG Partitioning for Parallel Workloads

Flavor name

GPU(s)

vRAM (GB)

vCPU(s)

RAM (GB)

Instance disk (GB)

Network (Gbit/s)

On-demand hourly (CHF)

On-Demand monthly (CHF)

Pre-Commit 1y monthly (CHF)

Pre-Commit 3y monthly (CHF)

a1g200d.64xlarge

4x H200

4x 141

192

768

960

≤ 20 (~12)

22.44

16'401.35

11'480.95

8'200.68

a1g200d.32xlarge

2x H200

4x 141

96

384

960

≤ 15 (~10)

11.22

8'361.64

5'845.15

4'175.82

a1g200d.16xlarge

H200

141

48

192

960

≤ 10 (~8)

6.01

4'468.02

3'127.62

2'234.01

GPU Power for Intense Workloads a1g6000d

This flavor is designed for demanding professional workloads and features the powerful NVIDIA RTX Pro 6000 Blackwell GPU with a generous 96 GB of GDDR7 ECC memory. This makes it ideal for complex 3D rendering, AI development directly on the workstation, and high-resolution real-time visualizations. For particularly demanding requirements, the configurable workstation instance a1g6000d.8xlarge is available. It combines 32 vCPUs, 128 GB of RAM, and 960 GB of NVMe SSD storage into an environment that requires no compromises, even for the most extensive workloads. These instances are designed for creative professionals, engineers, and researchers who require maximum graphics and computing power for their design, simulation, and AI applications. At its core is an AMD EPYC 9755 processor with an impressive 128 cores and a base clock speed of up to 4.1 GHz, ensuring reliable performance even under sustained full load.

Features

  • AMD EPYC 9755 128-Core Processor
  • NVIDIA RTX Pro 6000 Blackwell GPU with 96 GB vRAM
  • Up to 32 vCPUs at up to 4.1 GHz
  • Up to 128 GB RAM
  • Local NVMe instance block storage up to 960 GB and 250,000 IOPS
  • Up to 8 Gbit/s network bandwidth
  • Enhanced network performance via burst mode

Use Cases

  • LLM Inference with Long Context Windows
  • RAG Pipelines & Embedding Workloads
  • Agentic AI
  • LLM Training & Fine-Tuning
  • Medical Imaging
  • Financial Models & Fraud Detection
  • Scientific Simulations
  • MIG Partitioning

Flavor name

GPU

vRAM (GB)

vCPU(s)

RAM (GB)

Instance disk (GB)

Network (Gbit/s)

On-demand hourly (CHF)

On-Demand monthly (CHF)

Pre-Commit 1y monthly (CHF)

Pre-Commit 3y monthly (CHF)

a1g6000d.8xlarge

RTX Pro 6000

96

32

128

960

≤ 8 (~5)

3.54

2'632.43

1'842.71

1'316.22

GPU Power intense workloads p1gd

The latest flavor features an NVIDIA L4 Tensor Core GPU equipped with an impressive 24 GB of memory per GPU. For specific requirements, we offer a customized GPU instance (p1gd.4xlarge) equipped with 16 vCPUs, 64 GB RAM and 960 GB local SSD storage to handle even the most demanding applications and workloads with ease. These latest instances are specifically designed for highly complex AI, ML and graphics applications and offer an outstanding blend of performance, customization and expandability. They are based on the advanced AMD Ryzen 9 Zen4 processor, which reaches a base clock speed of up to 5.7 GHz.

Features

  • AMD Ryzen 9 Zen4 CPU
  • Nvidia L4 GPU
  • Up to 16 vCPU's with up to 5.7 GHz
  • Up to 64GB RAM
  • Local NVMe instance block storage up to 960 GB and 250,000 IOPS
  • Up to 250,000 kilobyte/s network bandwidth
  • Improved network performance through burst mode

Use Cases

  • AI
  • HPC
  • Video encoding
  • deep learning
  • Speech recognition
  • Compute-intensive cluster solutions
  • Scientific modeling
  • Data processing tasks with higher CPU requirements

Flavor name

GPU(s)

vRAM (GB)

vCPU(s)

RAM (GB)

Instance disk (GB)

Network (Gbit/s)

On-demand hourly (CHF)

On-Demand monthly (CHF)

Pre-Commit 1y monthly (CHF)

Pre-Commit 3y monthly (CHF)

p1gd.4xlarge

L4

24

16

64

960

≤ 2.5 (~2)

2.13

1'587.98

1'111.59

793.99

GPU memory

GPU memory is the memory that is available on a GPU device and can be used to temporarily store data. It is separate from the VM's main memory and has been specially developed for the higher bandwidth requirements of graphics-intensive workloads.

Local Storage

It is important to note that local SSD storage offers maximum storage performance, but data stored on a local SSD can be lost in the event of an SSD or system failure. We therefore strongly recommend that you make regular backups of your important data to avoid data loss. In addition, an instance cannot be started if the compute host is unavailable, which underlines the need for a robust backup and recovery strategy. Of course, as usual, you can also create volume-based instances with this flavor group that are not affected by the above limitations.

Zero-Disk g1

The flavor group g1 (Global Purpose Instances) offers a wide range of computing, storage and network resources. It does not include an instance disk (boot disk) and to use it, a bootable volume must first be created. In this way, the customer can flexibly select a volume type as a boot disk that meets their requirements.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)
  • Up to 24 vCPU's
  • Up to 96GB RAM
  • Up to 100'000 kilobyte/s network bandwidth
  • Distributed NVMe based storage solution with improved level of IOPS
  • Improved network burst performance for backend servers for EA
  • Volume block storage throughput depending on selected volume type
  • Volume block storage IOPS depending on selected volume type

Use Cases

  • Medium to large databases
  • Data processing tasks with higher demands on network bandwidth, storage size, IOPS and throughput
  • Microsoft SharePoint, cluster computing and other enterprise applications

Flavor-Name

vCPU(s)

RAM (GB)

Instance-Disk (GB)

On-Demand Stundensatz (CHF)

On-Demand Monatssatz (CHF)

Pre-Commit 1-Jahr Monatssatz (CHF)

Pre-Commit 3-Jahre Monatssatz (CHF)

g1.1c2m

1

2

0

0.07

49.22

34.45

24.61

g1.1c4m

1

4

0

0.08

57.78

40.45

28.89

g1.1c8m

1

8

0

0.10

74.90

52.43

37.45

g1.1c1m

1

1

0

0.06

44.94

31.46

22.47

g1.2c16m

2

16

0

0.18

130.54

91.38

65.27

g1.2c8m

2

8

0

0.13

96.30

67.41

48.15

g1.2c2m

2

2

0

0.09

70.62

49.43

35.31

g1.2c32m

2

32

0

0.27

199.02

139.31

99.51

g1.2c24m

2

24

0

0.22

164.78

115.35

82.39

g1.2c4m

2

4

0

0.11

79.18

55.43

39.59

g1.4c24m

4

24

0

0.28

207.58

145.31

103.79

g1.4c64m

4

64

0

0.51

378.78

265.15

189.39

g1.4c48m

4

48

0

0.42

310.30

217.21

155.15

g1.4c32m

4

32

0

0.33

241.82

169.27

120.91

g1.4c8m

4

8

0

0.19

139.10

97.37

69.55

g1.4c4m

4

4

0

0.16

121.98

85.39

60.99

g1.4c16m

4

16

0

0.23

173.34

121.34

86.67

g1.8c96m

8

96

0

0.81

601.34

420.94

300.67

g1.8c24m

8

24

0

0.39

293.18

205.23

146.59

g1.8c8m

8

8

0

0.30

224.70

157.29

112.35

g1.8c64m

8

64

0

0.62

464.38

325.07

232.19

g1.8c48m

8

48

0

0.53

395.90

277.13

197.95

g1.8c32m

8

32

0

0.44

327.42

229.19

163.71

g1.8c16m

8

16

0

0.35

258.94

181.26

129.47

g1.16c64m

16

64

0

0.85

635.58

444.91

317.79

g1.16c32m

16

32

0

0.67

498.62

349.03

249.31

g1.16c96m

16

96

0

1.04

772.54

540.78

386.27

g1.16c16m

16

16

0

0.58

430.14

301.10

215.07

g1.24c96m

24

96

0

1.27

943.74

660.62

471.87

g1.24c48m

24

48

0

0.99

738.30

516.81

369.15

Medium intense Workload m1

The Flavor Group m1 (Multi Purpose Instances) are ideal for less critical applications or test instances and offer a significantly lower performance of IOPS, disk throughput and network. It contains a defined instance disk.

Features

  • Intel(R) Xeon(R) processors (Skylake/Cascade Lake)
  • Up to 4 vCPU's
  • Up to 16GB RAM
  • Standard network burst performance
  • Storage throughput up to 150 MiB/s
  • Instance block storage IOPS up to 800
  • Instance block storage 10 GB to 120 GB
  • Up to 12,500 kilobyte/s network bandwidth
  • Distributed NVMe based storage solution with a standard level of IOPS

Use Cases

  • Web application services
  • Small databases
  • Project development and test environment
  • Data processing tasks with required standard memory size
  • Caching of fleets and operation of backend servers for EA

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

m1.small

1

2

50

0.05

40.45

28.31

20.22

m1.nano

1

0.5

10

0.04

27.61

19.32

13.80

m1.micro

1

1

20

0.04

31.67

22.17

15.84

m1.medium

2

4

80

0.10

75.11

52.58

37.56

m1.large

2

8

120

0.13

97.91

68.53

48.95

m1.xlarge

4

16

120

0.24

174.95

122.46

87.47

General Workload a1 (Region 2)

The Flavor Group a1 (Enterprise Performance Instances) offers a powerful combination of CPU, RAM, storage, and network resources, optimized for demanding applications with high computing power and memory requirements. It includes a defined instance disk with high capacity and fast NVMe-based storage solution. The ratio of vCPU to RAM is flexible and efficiently supports modern workloads.

Features

  • AMD EPYC processors (Turin)
  • Up to 64 vCPUs
  • Up to 256 GB RAM
  • Up to 500,000 kilobytes/s network bandwidth
  • Instance block storage up to 3,840 GB
  • Storage throughput up to 300 MiB/s
  • Instance block storage IOPS up to 3,000
  • Distributed NVMe-based storage solution with improved IOPS levels
  • Improved network burst performance

Use Cases

  • High-performance servers and compute-intensive applications
  • Large databases and analytics workloads
  • Applications with high network and storage throughput
  • Enterprise applications with high storage IOPS and network requirements

Flavor name

vCPU(s)

RAM (GB)

Instance disk (GB)

On-demand hourly rate (CHF)

On-Demand monthly rate (CHF)

Pre-Commit 1-year monthly rate (CHF)

Pre-Commit 3-year monthly rate (CHF)

a1.small

1

2

50

0.11

83.46

58.42

41.73

a1.medium

1

4

80

0.13

100.15

70.11

50.08

a1.large

2

8

120

0.20

148.09

103.66

74.04

a1.xlarge

4

16

240

0.32

240.54

168.38

120.27

a1.2xlarge

8

32

480

0.57

425.43

297.80

212.72

a1.4xlarge

16

64

960

1.07

795.22

556.66

397.61

a1.8xlarge

32

128

1920

2.06

1'534.81

1'074.37

767.40

a1.16xlarge

64

258

3840

4.05

3'013.98

2'109.78

1'506.99

GPU Power for Intense Workloads a1g6000d (Region 2)

This flavor is designed for demanding professional workloads and features the powerful NVIDIA RTX Pro 6000 Blackwell GPU with a generous 96 GB of GDDR7 ECC memory. This makes it ideal for complex 3D rendering, AI development directly on the workstation, and high-resolution real-time visualizations. For particularly demanding requirements, the configurable workstation instance a1g6000d.8xlarge is available. It combines 32 vCPUs, 128 GB of RAM, and 960 GB of NVMe SSD storage into an environment that requires no compromises, even for the most extensive workloads. These instances are designed for creative professionals, engineers, and researchers who require maximum graphics and computing power for their design, simulation, and AI applications. At its core is an AMD EPYC 9655P 96-core processor with an impressive 192 cores and a base clock speed of up to 4.5 GHz, ensuring reliable performance even under sustained full load.

Features

  • AMD EPYC 9655P 96-Core Processor
  • NVIDIA RTX Pro 6000 Blackwell GPU with 96 GB vRAM
  • Up to 32 vCPUs at up to 4.5 GHz
  • Up to 128 GB RAM
  • Local NVMe instance block storage up to 960 GB and 250,000 IOPS
  • Up to 5 Gbit/s network bandwidth
  • Enhanced network performance via burst mode

Use Cases

  • LLM Inference with Long Context Windows
  • RAG Pipelines & Embedding Workloads
  • Agentic AI
  • LLM Training & Fine-Tuning
  • Medical Imaging
  • Financial Models & Fraud Detection
  • Scientific Simulations
  • MIG Partitioning

Flavor name

GPU

vRAM (GB)

vCPU(s)

RAM (GB)

Instance disk (GB)

Network (Gbit/s)

On-demand hourly (CHF)

On-Demand monthly (CHF)

Pre-Commit 1y monthly (CHF)

Pre-Commit 3y monthly (CHF)

a1g6000d.8xlarge

RTX Pro 6000

96

32

128

960

≤ 8 (~5)

3.54

2'632.43

1'842.71

1'316.22