Operator · Boulogne-Billancourt, FR

Koyeb

French serverless platform running containers in microVMs on bare-metal servers, with per-second-billed GPU instances (RTX 4000 SFF Ada to 8x H200 and B200) and Tenstorrent accelerators; acquisition by Mistral AI announced February 2026.

GPU types10with a current offer

Current offers5454 readings on file · last 45 days

H100 from—per GPU-hour, on-demand

Entry price—cheapest GPU-hour listed

Countries517 named regions

Trust Score v0.1

82/100

Company & track record73
Infrastructure control100
Pricing & billing transparency85
Reliability signals100
Security & compliance47
Customer signal—

Scored Sep 29, 2026 from the provider's own pages. Dimensions not yet measured are excluded from the denominator. Criteria ↓

54 current offers →koyeb.com ↗

Trust Score · 5 of 6 dimensions measured

The score, criterion by criterion

Each point comes from something the provider publishes, with the page it was read on and the date. Anything not disclosed earns zero.

Company & track record

weight 15%

73 / 100

Founding year published	not disclosedkoyeb.com · Sep 29, 2026	0 / 4
Headquarters published	Koyebkoyeb.com · Sep 29, 2026	3 / 3
Legal entity named	Koyebkoyeb.com · Sep 29, 2026	4 / 4
Funding or ownership disclosed	Private; investors listed: Serena, Isai, Samsung Next, MongoDB, Acequia Capital, Kima Ventures, Plug & Play Ventures. Definitive agreement to join Mistral AI announced 17 Feb 2026, subject to closing conditions.koyeb.com · Sep 29, 2026	4 / 4

Infrastructure control

weight 15%

100 / 100

Capacity model published	operates or rents dedicated capacitykoyeb.com · Sep 29, 2026	10 / 10
Regions published	17 regions or countrieskoyeb.com · Sep 29, 2026	5 / 5

Pricing & billing transparency

weight 20%

85 / 100

Public prices without an account	yes (read on the pricing page)koyeb.com · Sep 29, 2026	5 / 5
Billing granularity published	per secondkoyeb.com · Sep 29, 2026	4 / 4
Egress policy published	metered: pricing page lists 1 TB/month included on Pro and Scale, then 0.02 USD/GB (EU/US) and 0.04 USD/GB (Asia); docs pricing FAQ states 100 GB free then 0.04 USD/GB and that outbound bandwidth is not charged yet (sources conflict)koyeb.com · Sep 29, 2026	4 / 4
Stopped-instance billing published	Paused Services are not charged. Scale-to-zero (public preview) puts idle GPU Services to sleep after an idle period (default 5 min, up to 6 h on Pro, 12 h on Scale). Billing starts when the Instance is created and the Deployment is in Starting status.koyeb.com · Sep 29, 2026	4 / 4
Public IPv4 policy published	not published (no static outbound IPs, shared regional IP pools; 'Advanced networking with direct IP' listed for Enterprise)koyeb.com · Sep 29, 2026	0 / 3

Reliability signals

weight 20%

100 / 100

Status page	https://status.koyeb.comkoyeb.com · Sep 29, 2026	6 / 6
SLA published	99.9% uptime on Scale plan, 99.99% on Enterprise (pro rata service credits on request)koyeb.com · Sep 29, 2026	6 / 6
SLA with a numeric uptime commitment	99.9% uptime on Scale plan, 99.99% on Enterprise (pro rata service credits on request)koyeb.com · Sep 29, 2026	3 / 3
Uptime history published	Status page 90-day figures on 2026-09-29: Control Panel 99.92%, API 99.96%, Build/Provisioning 99.75%, North America 99.98%, Europe 99.97%, APAC 99.91%koyeb.com · Sep 29, 2026	5 / 5
Observed availability (30 days)	not measured yet	—

Security & compliance

weight 15%

47 / 100

SOC 2	listed ('ISO27001 and SOC2 Certifications' under Enterprise plan); report type not published, Trust Center access on requestkoyeb.com · Sep 29, 2026	0 / 8
ISO 27001	listed under Enterprise plan features; certificate not readkoyeb.com · Sep 29, 2026	4 / 4
HIPAA or GDPR DPA	yes: public Data Processing Agreement under French law and GDPR; control plane on Google Cloud (Belgium); infrastructure subcontractors Equinix Metal, IBM Cloud, Scaleway, AWSkoyeb.com · Sep 29, 2026	3 / 3

Customer signal

weight 15%

not measured

Aggregated public reviews

not collected yet

—

provisional v0.1: evidence from the provider's own pages; dimensions not yet measured are excluded from the denominatorFull method →

Prices by GPU · 9 GPU types with a current offer

What Koyeb charges, against the market

Cheapest current on-demand true cost here, next to the market median for the same GPU. Negative means cheaper than the market.

GPU	Cheapest here	Market median	vs market	Current offers	Modes
H100	no on-demand	$3.17	—	4	serverless	H100 prices
A100 80GB	no on-demand	$1.92	—	8	serverless	A100 80GB prices
H200	no on-demand	$4.80	—	4	serverless	H200 prices
L40S	no on-demand	$1.24	—	1	serverless	L40S prices
RTX PRO 6000	no on-demand	$2.15	—	1	serverless	RTX PRO 6000 prices
B200	no on-demand	$7.50	—	1	serverless	B200 prices
RTX A6000	no on-demand	$0.54	—	1	serverless	RTX A6000 prices
L4	no on-demand	$0.80	—	1	serverless	L4 prices
RTX 4000 SFF Ada	no on-demand	—	—	1	serverless	RTX 4000 SFF Ada prices

All offers · 54 readings on file

Every Koyeb offer, with its receipt

Current offers (last 45 days) ranked by true cost, all rental modes, each with its read date. Older readings stay listed below, in grey.

GPU	Configuration	Mode	Region	Stock	List	True cost	Read
RTX 4000 SFF Ada	1× RTX 4000 SFF Ada6 vCPU	Serverless	fra	In stock	$0.50	$0.50	00:19 UTCfresh	Receipt
L4	1× L46 vCPU	Serverless	restricted plans only	Unavailable	$0.70	$0.70	00:19 UTCfresh	Receipt
RTX A6000	1× RTX A60006 vCPU	Serverless	na	In stock	$0.75	$0.75	00:19 UTCfresh	Receipt
L40S	1× L40S15 vCPU	Serverless	na	In stock	$1.20	$1.20	00:19 UTCfresh	Receipt
A100 80GB	1× A100 80GB15 vCPU	Serverless	na	In stock	$1.60	$1.60	00:19 UTCfresh	Receipt
A100 80GB	2× A100 80GB30 vCPU	Serverless	na	In stock	$1.60	$1.60	00:19 UTCfresh	Receipt
A100 80GB	4× A100 80GB60 vCPU	Serverless	na	In stock	$1.60	$1.60	00:19 UTCfresh	Receipt
A100 80GB	8× A100 80GB120 vCPU	Serverless	restricted plans only	Unavailable	$1.60	$1.60	00:19 UTCfresh	Receipt
A100 80GB	1× A100 80GB15 vCPU	Serverless	na	In stock	$2.15	$2.15	00:19 UTCfresh	Receipt
A100 80GB	2× A100 80GB30 vCPU	Serverless	na	In stock	$2.15	$2.15	00:19 UTCfresh	Receipt
A100 80GB	4× A100 80GB60 vCPU	Serverless	na	In stock	$2.15	$2.15	00:19 UTCfresh	Receipt
A100 80GB	8× A100 80GB120 vCPU	Serverless	restricted plans only	Unavailable	$2.15	$2.15	00:19 UTCfresh	Receipt
RTX PRO 6000	1× RTX PRO 600015 vCPU	Serverless	na	In stock	$2.20	$2.20	00:19 UTCfresh	Receipt
H100SXM	1× H100 SXM15 vCPU	Serverless	na	In stock	$2.50	$2.50	00:19 UTCfresh	Receipt
H100SXM	2× H100 SXM30 vCPU	Serverless	na	In stock	$2.50	$2.50	00:19 UTCfresh	Receipt
H100SXM	4× H100 SXM60 vCPU	Serverless	na	In stock	$2.50	$2.50	00:19 UTCfresh	Receipt
H100SXM	8× H100 SXM120 vCPU	Serverless	restricted plans only	Unavailable	$2.50	$2.50	00:19 UTCfresh	Receipt
H200	1× H20015 vCPU	Serverless	na	In stock	$3.00	$3.00	00:19 UTCfresh	Receipt
H200	2× H20030 vCPU	Serverless	na	In stock	$3.00	$3.00	00:19 UTCfresh	Receipt
H200	4× H20060 vCPU	Serverless	restricted plans only	Unavailable	$3.00	$3.00	00:19 UTCfresh	Receipt
H200	8× H200120 vCPU	Serverless	restricted plans only	Unavailable	$3.00	$3.00	00:19 UTCfresh	Receipt
B200	1× B20015 vCPU	Serverless	restricted plans only	Unavailable	$5.50	$5.50	00:19 UTCfresh	Receipt

22 fresh offers · prices per GPU-hour · true cost includes the add-ons the provider publishesOpen in the comparison tool →

True cost & billing · read Sep 29, 2026

What the hourly price does not say

Billing granularity, minimums, what a stopped machine costs, traffic, addresses, deposits and identity checks, as published by the provider.

Billing

Billing granularity: per second
Minimum charge: A paid plan is required for new sign-ups: Pro 29 USD/month (includes 10 USD compute) or Scale 299 USD/month (includes 100 USD compute); Enterprise from 1000 USD/month. Starter plan being removed for new users.
Stopped instance: Paused Services are not charged. Scale-to-zero (public preview) puts idle GPU Services to sleep after an idle period (default 5 min, up to 6 h on Pro, 12 h on Scale). Billing starts when the Instance is created and the Deployment is in Starting status.
Egress: metered: pricing page lists 1 TB/month included on Pro and Scale, then 0.02 USD/GB (EU/US) and 0.04 USD/GB (Asia); docs pricing FAQ states 100 GB free then 0.04 USD/GB and that outbound bandwidth is not charged yet (sources conflict)
Public IPv4: not published (no static outbound IPs, shared regional IP pools; 'Advanced networking with direct IP' listed for Enterprise)
Billing model: postpaid invoice
Minimum deposit: not published
Free credits: Pro plan includes 10 USD of compute per month, Scale 100 USD; startup program up to 30k USD in credits; /ai page states 'Try with $100 of free credit'.

Account

Payment: card
Sign-up: email, password, card
Identity check: card verification
Support: email, chat, community forum, slack (Scale)
Enterprise support: not published
SLA: 99.9% uptime on Scale plan, 99.99% on Enterprise (pro rata service credits on request)
Status page: status.koyeb.com ↗
Published uptime: Status page 90-day figures on 2026-09-29: Control Panel 99.92%, API 99.96%, Build/Provisioning 99.75%, North America 99.98%, Europe 99.97%, APAC 99.91%

Infrastructure

Regions: 17 named regions in 5 countries
FRA (Frankfurt) (Germany) · PAR (Paris) (France) · WAS (Washington · D.C.) (United States) · SFO (San Francisco · preview) (United States) · aws-us-east-1 (United States) · SIN (Singapore) (Singapore) · TYO (Tokyo) (Japan) · dal (Dallas · catalog API) (United States) · dsm (Iowa · catalog API) (United States) · mci (Kansas City · catalog API) (United States) · rdu (Durham · catalog API) (United States)
Countries: United States, Germany, France, Singapore, Japan
Platform: API, CLI, serverless, custom Docker images, persistent volumes, snapshotsSDK: Python and JavaScript (koyeb-sdk, documented for Sandboxes)Terraform: official
Storage: volume · None · Block volumes in public preview, documented as 1 to 10 GB, for was and fra regions, single-machine with no redundancy; price not published on pages read.ephemeral · included · Local disk included in each GPU instance type (160 GB to 2560 GB).
Migration out: Workloads are container images (Git build or any registry), so moving off Koyeb means redeploying the same image elsewhere. Volume data must be copied out from inside the running Service; snapshots live in Koyeb's remote store and no export path is documented. Configuration can be kept as Terraform or Pulumi code.

Security & compliance

SOC 2: listed ('ISO27001 and SOC2 Certifications' under Enterprise plan); report type not published, Trust Center access on request
ISO 27001: listed under Enterprise plan features; certificate not read
HIPAA: not published
GDPR / DPA: yes: public Data Processing Agreement under French law and GDPR; control plane on Google Cloud (Belgium); infrastructure subcontractors Equinix Metal, IBM Cloud, Scaleway, AWS
Other: Terms governed by French law, Paris courts.
Legal entity: Koyeb
Funding: Private; investors listed: Serena, Isai, Samsung Next, MongoDB, Acequia Capital, Kima Ventures, Plug & Play Ventures. Definitive agreement to join Mistral AI announced 17 Feb 2026, subject to closing conditions.

Read on the provider's own pages · 25 pages read · facts marked "not published" were looked for and not foundPages read ↓

Assessment · documents read Sep 29, 2026

Koyeb, in short

What stood out when we read the documents: strengths, weaknesses, who it suits, and what we could not find.

Strengths

Per-second billing with scale-to-zero on GPU Services (public preview), so idle inference endpoints stop costing after the idle period.
Full price list per instance, per second and per month visible without an account, including 2x/4x/8x A100, A100 SXM, H100 and H200.
Public instance catalog API (app.koyeb.com/v1/catalog/instances) exposes prices, regions and availability status.
Official Terraform provider, Pulumi provider, CLI, REST API with OpenAPI document, Python and JavaScript SDKs.
Only third-party cloud offering Tenstorrent Wormhole n300s accelerators, per Koyeb docs.
Published 99.9% (Scale) and 99.99% (Enterprise) SLA with service credits; public status page with uptime figures.
Saving plans advertised at up to 50% off on-demand for long-term reservations (terms on contact).

Weaknesses

New users must subscribe to a paid plan (Pro 29 USD/month minimum) since the Mistral AI announcement; the Starter tier is being removed.
Card-only payment, postpaid monthly; 29 USD pre-authorization hold and account validation that can take up to 3 business days.
Most GPU types are only available in North America per the catalog API; B200, L4, 4x H200 and all 8-GPU types show status RESTRICTED.
Persistent volumes and snapshots are in public preview, documented as 1 to 10 GB, single-machine, not redundant.
No spot or interruptible GPU tier; no published volume price.
H100 form factor (SXM or PCIe) and multi-GPU interconnect are not published.
Instance specs differ between the pricing page and the docs instance table (e.g. L40S 15 vCPU/64 GB vs 30 vCPU/92 GB).
Ownership transition: company announced joining Mistral AI and a gradual transition into Mistral Compute.

Ideal for

Teams deploying containerised inference APIs that benefit from autoscaling and scale-to-zero rather than long-running VMs.
Developers who want Git-push or Docker deploys with Terraform/Pulumi automation and do not need SSH into a GPU host.
Users evaluating Tenstorrent accelerators.

Not published

H100 form factor (SXM vs PCIe) not published on pricing, docs or catalog API
Founding year not stated as such
the about page only says the team 'set out' in 2021.
Volume and snapshot prices not published on pages read.
Saving-plan (reserved) prices and terms not published (up to 50% off, contact sales).
Tenstorrent n300s price not published (request access).
SOC 2 report type and ISO 27001 scope not verifiable: Trust Center requires an access request.
Egress: pricing page plan table and docs pricing FAQ give conflicting allowances and rates.
GPU interconnect (NVLink/NVSwitch) for multi-GPU instances not published.
Direct SSH on NVIDIA GPU instances not documented
SSH via TCP proxy is mentioned only for Tenstorrent.
Crypto, PayPal and wire transfer not mentioned
the FAQ lists card brands only.

Observations

Pricing page shows GPU instance prices per hour, per second and per month without an account; H100 at 2.50 USD/hour.

Banner on every page: 'Koyeb is joining Mistral AI to Build The Future of AI Infrastructure'.

N300s and 4x N300s show 'REQUEST ACCESS' instead of a price.

Plans table puts ISO27001/SOC2 and 99.99% SLA in Enterprise; GPU concurrency limit 20 on Pro, 40 on Scale.

FAQ on the pricing page mentions outbound bandwidth at 0.04 USD/GB while the plan table lists 0.02 USD/GB for EU/US.

Pages read

koyeb.com/pricing — GPU prices and specs, plans, egress, SLA tiers, compliance claims
koyeb.com/docs/reference/instances — Per-second/hour/month prices, instance specs (docs variant
app.koyeb.com/v1/catalog/instances?limit=100 — GPU prices per second/hour, regions, availability status
app.koyeb.com/v1/catalog/regions?limit=100 — Region and datacenter list
koyeb.com/docs/reference/regions — Core regions
koyeb.com/docs/faqs/pricing — Billing, card verification, payment methods, currency, taxes
koyeb.com/docs/reference/volumes — Persistent storage limits
koyeb.com/docs/reference/snapshots — Snapshots
koyeb.com/docs/run-and-scale/scale-to-zero — Idle behaviour for GPU services
koyeb.com/docs/reference/accounts — Sign-up methods
koyeb.com/docs/legal/sla — SLA mechanics, service credits
koyeb.com/docs/legal/data-processing-agreement — GDPR DPA, subcontractors
koyeb.com/docs/legal/terms — Legal entity, address, governing law
koyeb.com/docs/integrations/infrastructure-as-code/terraform — Official Terraform provider
koyeb.com/docs/integrations/infrastructure-as-code/pulumi — Pulumi provider
koyeb.com/docs/reference/api — REST API, OpenAPI
koyeb.com/docs/sandboxes/sandbox-quickstart — Python and JavaScript SDKs
koyeb.com/docs/hardware/nvidia-a100 — A100 PCIe vs SXM distinction
koyeb.com/docs/hardware/nvidia-h100 — H100 description (no form factor
koyeb.com/docs/hardware/tenstorrent-n300 — Accelerator offer, SSH via TCP proxy
koyeb.com/blog/koyeb-is-joining-mistral-ai-to-build-the-future-of-ai-infrastructure — Acquisition, paid-plan-only sign-up
koyeb.com/about — Team, history, investors
status.koyeb.com — Uptime figures
trust.koyeb.com — Compliance documents gated behind access request
koyeb.com/ai — Free credit mention

Questions about Koyeb

How does Koyeb bill?

Billing granularity: per second. Minimum charge: A paid plan is required for new sign-ups: Pro 29 USD/month (includes 10 USD compute) or Scale 299 USD/month (includes 100 USD compute); Enterprise from 1000 USD/month. Starter plan being removed for new users. Stopped instance: Paused Services are not charged. Scale-to-zero (public preview) puts idle GPU Services to sleep after an idle period (default 5 min, up to 6 h on Pro, 12 h on Scale). Billing starts when the Instance is created and the Deployment is in Starting status.

How trustworthy is Koyeb?

Its Trust Score is 82/100 (v0.1), read from its own documents. Strong on infrastructure control, reliability signals. Weaker on security & compliance. The score measures transparency and published guarantees, not customer satisfaction; every criterion and its source is listed above.

Where are Koyeb's data centres?

17 named regions in 5 countries: United States, Germany, France, Singapore, Japan.

How do you pay Koyeb, and are there credits?

Payment by card. Free credits: Pro plan includes 10 USD of compute per month, Scale 100 USD; startup program up to 30k USD in credits; /ai page states 'Try with $100 of free credit'.. Identity check: card verification.