Other providers get you started with an hourly rate that for most deep learning data scientists can spell trouble by the end of the month. With our service you'll always know what you're paying up front... no surprises.
We have pricing packages that let you plan for the entire life cycle of your projects. Purchase three, six or twelve months blocks of time and you can save up to 20% overall. That’s a huge savings.
Using a Graphcore Graphcloud instance with Cirrascale ensures no hidden fees with our flat-rate billing model. You pay one price without the worry of fluctuating bills like those at other providers. All pricing shown for PODs are per POD specified per month.
| IPU Instance | IPU Specs | Weekly Pricing | Monthly Pricing | Annual Pricing |
|---|---|---|---|---|
| Bow Pod 16 | 4x Bow-2000 | $3,250($19.35/hr*) |
$12,000($16.44/hr*) |
$9,600($13.15/hr*) |
| Bow Pod 64 | 16x Bow-2000 | $13,000($77.38/hr*) |
$48,000($65.75/hr*) |
$38,400($52.60/hr*) |
| Bow Pod 128 | 32x Bow-2000 | $26,000($154.76/hr*) |
$96,000($131.51/hr*) |
$76,800($105.21/hr*) |
| Bow Pod 256 | 64x Bow-2000 | $52,000($306.52/hr*) |
$192,000($263.01/hr*) |
$153,600($210.41/hr*) |
| Bow Pod 512 | 128x Bow-2000 | $104,000($619.05/hr*) |
$384,000($526.03/hr*) |
$307,200($420.82/hr*) |
| Bow Pod 1024 | 256x Bow-2000 | $208,000($1,238.10/hr*) |
$768,000($1,052.05/hr*) |
$614,400($841.64/hr*) |
| IPU-POD: Classic 16 | 4x IPU-M2000 | $2,437.50($14.51/hr*) |
$9,000($12.33/hr*) |
$7,200($9.86/hr*) |
| IPU-POD: Classic 64 | 16x IPU-M2000 | $9,750($58.04/hr*) |
$36,000($49.32/hr*) |
$28,800($39.45/hr*) |
| IPU-POD: Classic 128 | 32x IPU-M2000 | $19,500($116.07/hr*) |
$72,000($98.73/hr*) |
$57,600($78.90/hr*) |
| IPU-POD: Classic 256 | 64x IPU-M2000 | $39,000($232.14/hr*) |
$144,000($197.26/hr*) |
$115,200($157.81/hr*) |
| IPU-POD: Classic 512 | 128x IPU-M2000 | $78,000($464.29/hr*) |
$288,000($394.52/hr*) |
$230,400($315.62/hr*) |
| IPU-POD: Classic 1024 | 256x IPU-M2000 | $156,000($928.57/hr*) |
$576,000($789.04/hr*) |
$460,800($631.23/hr*) |
Introductory pricing below. All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. These prices represent blocks of dedicated cluster time for the chosen model. Additional system time is available at an hourly rate as needed.
| Model | Parameters | Tokens to Train to Chinchilla Point (B) | Cerebras AI Model Studio CS-2 Days to Train | Cerebras AI Model Studio Price to Train | Ready to Start |
|---|---|---|---|---|---|
| GPT3-XL | 1.3 | 26 | 0.4 | $2,500 | Sign-Up |
| GPT-J | 6 | 120 | 8 | $45,000 | Sign-Up |
| GPT-3 6.7B | 6.7 | 134 | 11 | $40,000 | Sign-Up |
| T-5 11B | 11 | 34* | 9 | $60,000 | Sign-Up |
| GPT-3 13B | 13 | 260 | 39 | $150,000 | Sign-Up |
| GPT NeoX | 20 | 400 | 47 | $525,000 | Sign-Up |
| GPT 70B | 70 | 1,400 | 85 | $2,500,000 | Sign-Up |
| GPT 175B | 175 | 3,500 | Contact For Quote | Contact For Quote | Contact For Quote |
* - T5 tokens to train from the original T5 paper. Chinchilla scaling laws not applicable.
Using a SambaNova instance with Cirrascale ensures no hidden fees with our flat-rate billing model. You pay one price without the worry of fluctuating bills like those at other providers. Pricing shown below is per instance for the total time stated.
| Instance | Weekly Rate | Monthly Rate | Annual Rate | 3-Year Rate |
|---|---|---|---|---|
| DataFlow-as-a-Service | Contact for Pricing | Contact for Pricing | Contact for Pricing | Contact for Pricing |
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. All pricing shown for servers are per server per month.
| GPU | Processor Specs |
System RAM | Local Storage | Network | Hourly Equiv. Pricing* |
Monthly Pricing | 6-Month Pricing | Annual Pricing |
|---|---|---|---|---|---|---|---|---|
| 8X NVIDIA H100 (Standalone) |
Dual 48-core | 2TB | (1) 960GB NVMe (4) 3.84TB NVMe |
25Gb Bonded | $34.25*($4.28/GPU/hr*) |
$24,999 | $22,499 | $19,999 |
| 8X NVIDIA H100 (16-Node+ Clustered) |
Dual 48-core | 2TB | 960GB NVMe (4) 3.84TB NVMe |
4x400Gb InfiniBand | As low as $2.49 per GPU hour* |
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. All pricing shown for servers are per server per month.
| GPU | Processor Specs | System RAM | Local Storage | Network | Hourly Equiv. Pricing* | Monthly Pricing | 6-Month Pricing | Annual Pricing |
|---|---|---|---|---|---|---|---|---|
| 8X NVIDIA A100 (80GB) | Dual 32-core | 1TB | 3.84TB NVMe | 25Gb Bonded | $26.03 | $18,999 | $17,159 | $15,319 |
| 4X NVIDIA A100 (80GB) | Dual 24-core | 1TB | 1.92TB NVMe | 25Gb Bonded | $13.70 | $9,999 | $8,999 | $7,999 |
| 8X NVIDIA A100 (40GB) | Dual 32-core | 1TB | 3.84TB NVMe | 25Gb Bonded | $22.96 | $16,759 | $15,079 | $13,559 |
| 8X NVIDIA RTX A6000 | Dual 32-core | 512GB | 3.84TB NVMe | 25Gb Bonded | $8.97 | $6,549 | $5,899 | $5,239 |
| 8X NVIDIA RTX A5000 | Dual 10-core | 256GB | 1TB NVMe | 25Gb Bonded | $4.11 | $2,999 | $2,699 | $2,399 |
| 8X NVIDIA RTX A4000 | Dual 10-core | 256GB | 1TB NVMe | 25Gb Bonded | $2.74 | $1,999 | $1,799 | $1,599 |
| 8X NVIDIA A40 | Dual 32-core | 512GB | 3.84TB NVMe | 25Gb Bonded | $11.53 | $8,419 | $7,579 | $6,819 |
| 8X NVIDIA A30 | Dual 32-core | 512GB | 3.84TB NVMe | 25Gb Bonded | Call | Call | Call | Call |
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. All pricing shown for servers are per server per month.
| GPU | Processor Specs | System RAM | Local Storage | Network | Hourly Equiv. Pricing* | Monthly Pricing | 6-Month Pricing | Annual Pricing |
|---|---|---|---|---|---|---|---|---|
| 8X Tesla V100 (DGX-1) |
Dual Intel Xeon E5-2698 v4 | 512GB | 4X 1.92TB NVMe | 25Gb Bonded | $20.55 | $14,999 | $13,499 | $11,999 |
| 8X Tesla V100 | Dual Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $17.81 | $12,999 | $11,699 | $10,399 |
| 8X Tesla V100 | AMD EPYC 7402P | 256GB | 1TB NVMe | 25Gb Bonded | $15.34 | $11,199 | $10,079 | $8,959 |
| 4X Tesla V100 | Dual IBM POWER 9 | 256GB | 1.6TB NVMe 2TB SATA |
25Gb Bonded | $3.29✝ | $2,399✝ | $2,399✝ | $2,399✝ |
| 4X Tesla V100 | Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $10.27 | $7,499 | $6,749 | $5,999 |
| 8X Tesla P100 | Dual Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $9.59 | $6,999 | $6,299 | $5,599 |
| 4X Tesla P100 | Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $5.48 | $3,999 | $3,599 | $3,199 |
| 4X Tesla P100 | Dual IBM POWER 8 | 256GB | 1TB NVMe / 1TB SATA |
25Gb Bonded | $1.78✝ | $1,299✝ | $1,299✝ | $1,299✝ |
✝Supplies limited. Pricing limited for quantity available on hand.
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. All pricing shown for servers are per server per month.
| GPU | Processor Specs | System RAM | Local Storage | Network | Hourly Equiv. Pricing* | Monthly Pricing | 6-Month Pricing | Annual Pricing |
|---|---|---|---|---|---|---|---|---|
| 8X Quadro RTX 8000 | Dual Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $11.77 | $8,599 | $7,739 | $6,879 |
| 8X Quadro RTX 6000 | Dual Intel Xeon E5-2630 v4 | 256GB | 1TB NVMe | 25Gb Bonded | $9.45 | $6,899 | $6,209 | $5,519 |
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested. All pricing shown for servers are per server per month.
| GPU | Processor Specs | System RAM | Local Storage | Host Card | Hourly Equiv. Pricing* | Monthly Pricing | 6-Month Pricing | Annual Pricing |
|---|---|---|---|---|---|---|---|---|
| 2X Quadro RTX 8000 | Intel® Xeon® W-2145 (3.7GHz) | 128GB | 2TB NVMe | Teradici PCoIP Host Card 2240*** | $3.83 | $2,799 | $2,519 | $2,239 |
*** Connection client not included. Zero Clients are available starting at $79 per month.
All pricing below is based on Cirrascale's No Surprises billing model. There are no hidden fees and discounts may apply for long-term commitments depending on the service requested.
| Solution | Capacity | Price / GB / Month |
|---|---|---|
| NVMe Hot-Tier Storage | 50TB or Greater | $0.20 |
| Object Storage | Less Than 50TB | $0.04 |
| Object Storage | 50TB - 2PB | $0.02 |
| Object Storage | 2PB or Greater | $0.01 |
Ready to take advantage of our flat-rate monthly billing, no ingress/egress data fees, and fast multi-tiered storage?
Hey Soldier welcome back signin now there is lot of new stuff waiting for you
Login with Facebook Login with Twitter
If you are interested in using the Cerebras AI Model Studio, part of the Cerebras Cloud @ Cirrascale, please fill out our sign-up form. Please completely fill out the form so that we can ensure an accurate review of your request.
Once received, we'll review you information and have one of our onboarding representatives give you a call to schedule your service.
If you are interested in gaining access to SambaNova Dataflow-as-a-Service, please fill out our sign-up form. Please completely fill out the form so that we can ensure an accurate review of your request.
Once received, we'll review you information and have one of our onboarding representatives give you a call to schedule your service.