Google Cloud vs Hikube — Direktvergleich

Interaktive Fassung: https://comparison.cloud/vergleich/google-vs-hikube

Vendor-neutraler Direktvergleich von Google Cloud und Hikube: Services, Rechenzentren, Kosten und Souveränität — datenbasiert auf comparison.cloud.

Management Summary (Stand 2026-10)

Google Cloud positioniert sich als technologisch führende Plattform für daten- und KI-getriebene Workloads, mit Spitzenwerten in Big Data (10/10) und AI (9/10), bleibt jedoch aufgrund der US-Jurisdiktion mit einem SEAL Score von 4,3/10 für souveränitätskritische Szenarien eingeschränkt. Hikube ist ein Schweizer Anbieter, der durch ein architektonisch integriertes 3-AZ-Design über Genf, Gland und Luzern eine hohe operative Resilienz (10/10) und technologische Souveränität (9/10) bietet. Während Google Cloud mit einem breiten Portfolio aus 100+ Services und globaler Reichweite punktet, setzt Hikube auf einen fokussierten Open-Source-Stack mit geringem Lock-in und attraktiven GPU-Preisen (Cost Rating 9,7/10). Für Unternehmen, die maximale Innovationskraft und globale Skalierbarkeit priorisieren, ist Google Cloud die geeignete Wahl. Für Organisationen mit hohem Souveränitätsbedarf, Fokus auf Schweizer Infrastruktur und spezifischen GPU-Workloads bietet Hikube eine spezialisierte Alternative.

Google Cloud: Stärken

Google Cloud: Zu bedenken

Hikube: Stärken

Hikube: Zu bedenken

Google Cloud wählen, wenn:

Hikube wählen, wenn:

Häufige Fragen: Google Cloud vs Hikube

Was ist der Unterschied zwischen Google Cloud und Hikube?

Google Cloud ist ein globaler Hyperscaler mit einem breiten Portfolio (100+ Services) und Spitzenwerten in Big Data (10/10) und AI (9/10), operiert jedoch unter US-Jurisdiktion (SEAL 4,3/10). Hikube ist ein Schweizer Anbieter mit einem fokussierten Portfolio (20+ Services), der durch ein 3-AZ-Design in der Schweiz (Hochverfügbarkeit 10/10) und hohe technologische Souveränität (9/10) punktet, aber bei strategischen Zusatzdiensten wie Big Data (3/10) eingeschränkt ist.

Welcher Anbieter ist souveräner?

Hikube ist deutlich souveräner mit einem SEAL Score von 6,9/10 (69 %) im Vergleich zu Google Cloud mit 4,3/10 (43 %). Dies liegt vor allem an der Schweizer Jurisdiktion und dem Open-Source-Stack von Hikube (Technologische Souveränität 9/10 vs. 6/10 bei Google). Google Cloud bietet zwar technische Souveränitäts-Features wie Confidential Computing, bleibt aber jurisdiktionell an die USA gebunden (SOV-2: 2/10).

Welcher Anbieter ist günstiger?

Hikube hat ein leicht besseres Gesamt-Cost Rating (6,4/10) als Google Cloud (5,9/10). Besonders bei GPU-Instanzen ist Hikube mit 9,7/10 deutlich günstiger als Google Cloud (3,6/10). Bei CPU-Instanzen liegt Hikube ebenfalls vorn (6,2/10 vs. 3,9/10). Allerdings ist Google Cloud bei Generative AI-Token-Preisen mit 10,0/10 sehr wettbewerbsfähig. Die tatsächlichen Kosten sind stark workload-abhängig.

Welcher Anbieter bietet bessere Hochverfügbarkeit?

Hikube bietet mit einem Score von 10/10 (Ø: 6,9) die höchste Hochverfügbarkeit im Vergleichsfeld, da jede Workload aktiv-aktiv über drei Schweizer Rechenzentren (Genf, Gland, Luzern) läuft. Google Cloud erreicht 9/10 und bietet 130 Availability Zones weltweit, was für globale Verteilung vorteilhaft ist, aber nicht die gleiche integrierte Multi-Site-Resilienz in einer Region bietet.

Für welche Workloads ist Google Cloud besser geeignet als Hikube?

Google Cloud ist besser geeignet für datenintensive Workloads wie Big Data (10/10), komplexe KI-Modelle (9/10) und globale Microservices-Architekturen (Containerisierung 10/10). Hikube ist dagegen besser für Workloads, die auf Schweizer Infrastruktur, hohe Resilienz ohne eigene DR-Planung und kostengünstige GPU-Nutzung (9,7/10) angewiesen sind, aber keine umfangreichen Big-Data- oder Collaboration-Tools (1/10) benötigen.

Management Summary

ServiceGoogle CloudHikube
Count of Services100+20+
Cost Rating5.9/106.4/10
Saving Plan57%–
Datacenter130 zones (in 43 regions; Google publishes no data-center count)3
Datacenter ModelHybrid (Google-owned or leased data centers; a region usually spans 3+ data centers)Colocation
Zone SeparationSeparate data centers (3+ physical data centers per region; Stockholm, Osaka, Montréal: 1–2)Separate buildings
Server ProductionCustom design–
Countries291
SEAL Score4.3/10 (43%)6.9/10 (69%)
USPLeading in data analytics, AI, and Kubernetes, adjustable vCPU/RAM ratio3 AZ by design: every workload runs active-active across three Swiss datacenters. VM volumes and Kubernetes are replicated by the platform.
RecommendationRecommended for data-driven organizations and developers focused on innovationRecommended for Swiss organisations and regulated mid-sized companies that want built-in HA resilience without designing it themselves.

Sovereign Cloud

ServiceGoogle CloudHikube
InterpretationGermany Data Boundary by T-Systems: resources only in Frankfurt (europe-west3), CMEK mandatory, only external keys (Cloud EKM) allowedSwiss operator (ownership chain not publicly evidenced); compute and storage run in three Swiss datacenters, transfers abroad only with consent except in security emergencies or legal obligations. The buildings are colocation sites run by third parties, sub-processor locations are not published, and the platform only opened in 2025
SEALSEAL 1SEAL 3
ResidencyUSSwitzerland
Shared ResourcesGermany Data Boundary by T-Systems (Sovereign Controls by Partners folder; europe-west3 only)Cloud Instances
Dedicated ResourcesSole-Tenant Nodes–
Dedicated CloudGoogle Distributed CloudHikube (on Prem) on Request
EncryptionCloud HSMVault as a Service (secrets management in the Swiss tenant)
Ownership & controlUltimate parent Alphabet Inc. (Delaware, USA) (Page/Brin hold 52.7% of voting power; 10-K FY2025, as of 2025-12-31)Hidora SA, Petit-Lancy, Switzerland (not EU), CHE-286.910.173 (commercial register GE/Zefix; AG since 2016, capital CHF 144,750 in restricted registered shares)
EU governance & oversightNo independent EU body with veto or oversight rightsNo independent board or advisory body with veto/oversight rights (checked: about, imprint, security & compliance, GTC)
Third-country access riskGoogle LLC subject to the CLOUD Act and FISA 702Hidora SA, Switzerland (not EU) (ownership chain not publicly evidenced)
Contracting entity & jurisdictionEMEA contracting entity Google Cloud EMEA Limited (Dublin) (FR/IT/PL local entities)Contracting party HIDORA SA, Petit-Lancy; Swiss law, Geneva courts (GTC October 2026 clause 21; CHE-286.910.173; Switzerland (not EU))
Transparency on authority requestsTransparency Report counts requests on Enterprise Cloud customers (whitepaper 'Government Requests for Cloud Customer Data' (02/2022); excl. US national security requests)No transparency report on government requests (checked: security & compliance, CLOUD Act guide (2026-06-10, upd. 2026-08-21), resources, blog)
Data residency incl. backups/DRCustomer Data at rest only in the chosen region (General Service Terms §1 'Data Location'; backup replication in-country only)Operated from Swiss DCs: Geneva (Plan-les-Ouates), Gland, Lucerne (Hikube service agreement October 2026, clause 3 and Annex A; Switzerland (EU adequacy decision))
Metadata & support data residencyNo commitment for telemetry, billing, support data (standard operations)Ancillary processing 'may involve subcontractors' (service agreement October 2026 clause 3: anti-DDoS, monitoring, support)
AI services: processing & trainingNo training on customer data without permission (Gemini Enterprise Agent Platform (ex Vertex AI))Not applicable: no managed AI service (GPU as a Service only, 6 NVIDIA cards L4–H200 on VMs/node groups; 'no opaque ML PaaS')
Dedicated / segregated environmentGDC air-gapped: infrastructure and services on your premises (no connectivity to Google Cloud required, private isolated environment, no internet access)Managed Kubernetes control plane dedicated to your cluster (own API server, own etcd, 'shared with nobody'; workers in own tenant)
EU operations & supportAssured Workloads: Customer Data in use processed only in the EU (Service Terms §40, EU region; access by Google personnel outside the region can be prevented)Support by Hidora SA engineers in Geneva (support page 2026-08-23; business days 08:00–18:00 CET, 24×7 infrastructure on-call)
Technical operator lock-outConfidential VM: keys kept in dedicated hardware (AMD SEV/SEV-SNP, Intel TDX, NVIDIA CC; inaccessible to the hypervisor)No confidential computing, no attestation, no HYOK
Disconnected / autonomous operationGDC air-gapped 'does not require connectivity to Google Cloud' (on-prem on OEM hardware, software via portable media, FIPS 140-2 Rocky Linux)Control plane operated exclusively in Switzerland (guide 2026-07-15, upd. 2026-09-07: Geneva, Gland and Lucerne; Switzerland (not EU))
Hardware origin & manufacturingGoogle designs server boards, networking and the Titan chip (Infrastructure Security Design Overview; component vendors vetted)Only the GPU vendor disclosed: NVIDIA (six cards L4 24G, L40S 48G, A100 80G, RTX 6000 Pro 96G, H100 80G, H200 141G)
Critical software dependenciesStack is proprietary Google (USA) (Borg, ALTS, central identity service, KVM-based hypervisor; disclosed in design docs, not mitigated)Partially disclosed: Talos Linux, Cilium, Grafana, VictoriaMetrics (Managed Kubernetes page; VictoriaLogs, cert-manager)
Open-source share & patch authorityGoogle open source incl. Kubernetes, Istio, Knative, Go (TensorFlow, Angular; opensource.google)Managed OSS services: OpenBAO, Harbor/Trivy, Grafana, MariaDB (release notes Hikube v1.1.2 (19 Mar 2026); OpenBAO = open-source fork of HashiCorp Vault; RabbitMQ)
Portability & exitEU Data Act Terms: no switching charges, free transfer on exit (General Service Terms §16, EEA)Reversibility: return 'in a reasonable standard format' (GTC Oct 2026 cl. 9.2; content kept thirty (30) days after termination, then deletion (Annex D cl. 11))
IaC & automationAt least 12 months' notice before discontinuing a Service (Google Cloud Terms section 1.4(e))Public REST/JSON API with OpenAPI 2.0 specification (api.hikube.cloud: VMs, disks, S3 buckets, Kubernetes, VPC, databases, backups, GPU, pricing)
Key sovereignty (BYOK/HYOK, EU HSM)Cloud EKM (HYOK): key material stays in the external EKM ('never cached or stored' in Google Cloud; partners Fortanix, Futurex, Thales)Vault as a Service in your Swiss tenant, operated by Hidora ('without US KMS or US Vault'; 'no self-operated HSM')
Certifications & scopeBSI C5:2020 attestation per Google with service list (Compute Engine, GKE, BigQuery etc.; report, Type 1/2 and region scope login-only)ISO 27001 'certified by SQS' per the provider ('the PDF is not online'; attestation, scope annex, validity date on request only)
Public audit evidenceCompliance Reports Manager: ISO/IEC certificates, SOC reports (self assessments; 'Select resources may require sign-in')Public only: GTC, Hikube contract, privacy notice, status page
EU labels & schemesEU Cloud CoC: Google LLC, Level 2 (Adherence ID 2020LVL02SCOPE015; partly covered by third-party audits)No EUCS, Gaia-X, EU Cloud CoC, CISPE, SecNumCloud or Cloud III SEAL (no BSI C3A; Swiss provider (not EU))
Energy & CO₂ transparencyPUE per campus published quarterly (Q2/2026 TTM: Hamina 1.10, St. Ghislain 1.08, Eemshaven 1.07, Dublin 1.08; fleet 1.09 (2025))'No PUE published'; 100% renewable power at three sites (infrastructure page; Gland approx. 6.9 g CO₂-eq/kWh, Geneva waste heat up to 4.4 MW (CAD-ZIPLO))
Climate commitmentsISO 50001 at five of six Google-owned data centres in Europe (Environmental Report 2025; limited assurance 2025 (FY2024))No own climate targets (Hidora/Hikube not in the Climate Neutral Data Centre Pact public register)

SEAL Framework

ServiceGoogle CloudHikube
Sovereignty Score4.3/10 (43%) Open technology helps; US jurisdiction limits sovereignty6.9/10 (69%) Swiss operator (ownership chain not publicly evidenced) on an open-source stack in colocation sites; own compliance certification ISO 27001
SOV-1 Strategic (15 %)US parent (Alphabet); sovereignty via partners (S3NS, T-Systems)Founder-led Swiss company Hidora SA; ownership chain not publicly evidenced; small provider, platform launched in 2025
SOV-2 Legal & Jurisdictional (10 %)Subject to US CLOUD Act & FISA 702Swiss jurisdiction (nFADP), no US parent; EU adequacy decision for Switzerland covers transfers from the EU
SOV-3 Data & AI (10 %)Sovereign controls & external key managementData in three Swiss datacenters; transfers abroad only with consent, except in security emergencies or legal obligations; dedicated GPUs, but no managed AI services
SOV-4 Operational (15 %)Partner-operated sovereign models; global ops otherwiseOperated and supported by Swiss-based engineers; buildings (power, cooling, physical access) are run by third-party site operators
SOV-5 Supply Chain (20 %)Global proprietary infrastructure supply chainColocation in three third-party buildings (Stack Infrastructure, backed by US investor IPI Partners, in Geneva and Gland; EWL in Lucerne); hardware sourced on the global market
SOV-6 Technology (15 %)Strong open source (Kubernetes); moderate lock-inOpen-source, Kubernetes-native stack (Talos Linux, Cilium, OpenBAO, upstream PostgreSQL/MariaDB/Redis/Kafka), S3 API, kubectl/Helm/GitOps, VMDK/QCOW2/ISO import; hypervisor and storage layer not disclosed
SOV-7 Security & Compliance (10 %)Broad certifications; SecNumCloud via partner S3NSISO 27001 (operator ISMS, certified by SQS; certificate not published online); no C5, SOC 2, ISO 27017/27018; SLA 99.99% per service published (GPU passthrough and preview services excluded); RPO/RTO not published (Hidora states < 5 min, publication announced)
SOV-8 Environmental (5 %)24/7 carbon-free energy goal by 2030100% renewable power at all three sites, heat reuse in Geneva and Lucerne; no PUE published, ISO 14001/50001 are building certificates

Availability & Locations

ServiceGoogle CloudHikube
Presence DACHGermanySwitzerland
Presence EuropeBelgium St. GhislainSwitzerland Geneva (Plan-les-Ouates), Gland, Lucerne
Presence North AmericaUSA Iowa, Virginia, Oregon, Texas …–
Presence South AmericaBrazil São Paulo–
Presence Asia-PacificTaiwan Changhua–
Presence Middle East & AfricaQatar Doha–
Availability Zones1303
Multi-AZ✓ verfügbar✓ verfügbar
Regions431
Countries291
Continents61
Point of Presence100+–
Dedicated Backbone✓ verfügbar–
DDoS✓ verfügbar–

Compute

ServiceGoogle CloudHikube
Shared InstanceCompute Engine 1vCPU-416vCPUCloud Instances s1 / u1 / m1, 1vCPU-64vCPU
Dedicated InstanceSole-Tenant Nodes–
GPU InstanceCompute Engine A5X, A4X Max, A4X, A4, A3, A2, G4, G2GPU Cloud NVIDIA L4, L40S, A100, RTX 6000 Pro, H100, H200
AI AcceleratorCloud TPU7x (Ironwood) 7th gen, 9,216 chips per pod, 192 GiB HBM per chip; JAX and PyTorch–
Windows ServerCompute EngineWindows Server licence (incl. 2025, per vCPU)
ServerlessCloud Run–
ARM InstanceAxion (C4A, N4A)–
Confidential ComputingConfidential VMs–

Hot Storage

ServiceGoogle CloudHikube
Block StoragePersistent DiskBlock Storage 3x Replicated NVMe, sync or async across 3 datacenters
AssetsCloud StorageSwiss S3 Object Storage (S3 API, WORM)
Managed NetAppCloud NetApp Volumes–
FileFilestore–

Cold Storage

ServiceGoogle CloudHikube
VM BackupColdline StorageBackup as a Service
Volume BackupDisk snapshots standard and archive snapshotsVelero (Kubernetes add-on)
ArchiveColdline Storage–
TapeArchive Storage–
Disaster RecoveryBackup and DR Service3-datacenter replication by design (restart on the remaining sites)

Managed Databases

ServiceGoogle CloudHikube
MySQLCloud SQLManaged MariaDB
PostGreCloud SQLManaged PostgreSQL (CloudNativePG)
MS SQLCloud SQL for SQL Server–
OracleOracle Database@Google Cloud–
Document-OrientedMongoDB AtlasManaged MongoDB
Distributed Wide-ColumnBigtableManaged ClickHouse
TSDBBigQuery–
CacheMemoryStoreManaged Redis (Sentinel)
Search EngineCloud Search–
Knowledge GraphSpanner Graph–
VectorVector Search Gemini Enterprise Agent Platform (formerly Vertex AI)Managed Qdrant

Network & Security

ServiceGoogle CloudHikube
NetworkVirtual Private CloudVPC & Subnets
Public IPPublic IPPublic IPv4
GatewayAPI Gateway–
Load BalancerCloud Load BalancingManaged Ingress (ingress-nginx, Gateway API)
Public DNSCloud DNS–
Private DNSCloud DNS - Private Zone–
FirewallCloud ArmorPer-VM firewall (allowed inbound ports on the public IP; SSH 22 default, HTTP/HTTPS optional)
Private ConnectivityCloud VPNSite-to-site VPN (on request)
Dedicated ConnectivityCloud Interconnect–
DDoS ProtectionCloud Armor–
CDNCloud CDN–

Orchestration

ServiceGoogle CloudHikube
Container RegistryArtifact RegistryRegistry as a Service
OrchestrationGoogle Kubernetes Engine(GKE)Managed Kubernetes

Virtualization

ServiceGoogle CloudHikube
HypervisorKVM no direct accessNot disclosed (VM layer operated by Hikube; VMDK / QCOW2 / ISO import)
Instance as a ServiceCompute EngineCloud Instances
Managed Virtualization PlatformsGoogle Cloud VMware Engine–
Bare Metal VirtualizationBare Metal Solution–
Edge VirtualizationGoogle Distributed Cloud–
Workload-specific VirtualizationSAP RISE on Google Cloud–

Management

ServiceGoogle CloudHikube
Identity ManagementCloud IdentityOrganizations & Projects (isolated projects with quotas, members and roles; API keys per project)
IaCInfrastructure Manager managed TerraformHikube public API (REST/JSON, OpenAPI 2.0 spec)
Audit TrailCloud Audit Logs–
Command Linegcloud CLIkubectl / Helm
MCP ServerGoogle Cloud remote MCP servers BigQuery, Cloud Run, Cloud SQL, GKE, Spanner …–
CO2 CalculatorCarbon Footprint–
CollaborationGoogle Workspace–

Big Data

ServiceGoogle CloudHikube
DashboardsCloud Monitoring dashboardsGrafana (Monitoring Agents add-on, per cluster)
Data StreamingDatastreamManaged Kafka
MessagingCloud Pub/SubManaged RabbitMQ
Business AnalyticsLooker–
Big Data ProcessingManaged Service for Apache Spark formerly Dataproc; clusters and serverless–
Big Data PlattformBigQueryManaged ClickHouse (OLAP)
Log MonitoringCloud MonitoringMonitoring Agents add-on

AI & Machine Learning

ServiceGoogle CloudHikube
AssistantDialogflow–
TranslationTranslation AI–
Audio AnalysisSpeech-to-Text API Chirp 3 transcription (V2 API), diarization, language detection–
Computer VisionVision AI–
NLPNatural Language API–
Image GenerationGemini Enterprise Agent Platform Gemini Image (Nano Banana): Nano Banana 2.1, Gemini 3.1 Flash Image, 3 Pro Image, Nano Banana 2 Lite–
Machine LearningGemini Enterprise Agent Platform formerly Vertex AIGPU as a Service (dedicated NVIDIA GPUs on VMs or Kubernetes node groups)
Document IntelligenceDocument AI–
Large Language ModelsGemini Enterprise Agent Platform Model Garden, 200+ models: Gemini, Claude, GPT, Grok, Mistral, Meta, open models–
EmbeddingsGemini Enterprise Agent Platform–
Serverless InferenceStandard PayGo pay-per-token, spend-based usage tiers (Flash up to 50M TPM, Pro up to 10M TPM)–
Batch InferenceGemini batch inference −50 % vs. real-time, ≤200k requests / 1 GB per job, most jobs done in 24 h–
Dedicated InferenceProvisioned Throughput reserved GSU capacity; 1 wk/1/3 mo/1 yr terms; 1-yr −26 % vs. 1-month–
Service TiersPriority PayGo premium per-token rate, more consistent performance; global, us & eu endpoints–
Fine-TuningGemini supervised fine-tuning 3.5 Flash, 3.5/3.1 Flash-Lite (+2.5); tuning & serving on us/eu multi-region–
GuardrailsModel Armor prompt injection/jailbreak, harmful content, sensitive data (SDP), malicious URLs–
AI Data PolicyNo training on customer data also for pre-GA models; ZDR needs opt-outs (caching, abuse logging, grounding)–
AI RouterApigee AI gateway cross-provider model routing, automatic fallback, semantic caching, token quotas–
AI AgentsAgent Runtime managed hosting for ADK, LangChain/LangGraph, LlamaIndex, CrewAI or any container–
RAG & GroundingRAG Engine managed RAG corpora; GA in europe-west3 (Frankfurt) and europe-west4 (NL)–
Prompt CachingImplicit caching (default on) −90 % on cached tokens, no storage fee, max. 24 h; disable for zero data retention–
Evaluation & ObservabilityGen AI evaluation service adaptive rubrics for model migration, prompt editing, tuning; EU regions (no Frankfurt)–
Video & Music GenerationGemini Enterprise Agent Platform Veo 3.1 / 3.1 Fast: text/image-to-video with sound, 4–8 s, up to 4K; us-central1 only; 3.1 Lite (Preview)–
Sovereign / On-Prem AIGemini on GDC air-gapped Gemini 3.5 Flash on NVIDIA HGX B300, entirely inside the air-gapped perimeter–

LLM Models

ServiceGoogle CloudHikube
Qwen3-235B-A22B-Instruct-2507 (Alibaba Cloud · open weights)✓ verfügbar · 262k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
Qwen3-Coder-480B-A35B-Instruct (Alibaba Cloud · open weights)✓ verfügbar · 262k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
Qwen3-Next-80B-A3B-Thinking (Alibaba Cloud · open weights)✓ verfügbar · 262k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
Claude Fable 5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Fable 5.1 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Haiku 4.5 (Anthropic · proprietary)✓ verfügbar · 200k Kontext (retirement not before 07.12.2026)–
Claude Haiku 5.5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Opus 4.5 (Anthropic · proprietary)✓ verfügbar · 200k Kontext (retirement not before 24.11.2026)–
Claude Opus 4.6 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Opus 4.7 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Opus 4.8 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Opus 5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Opus 5.5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Sonnet 4.5 (Anthropic · proprietary)✓ verfügbar · 200k Kontext (deprecated; EOL 30.11.2026 (successor Sonnet 4.6+))–
Claude Sonnet 4.6 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Sonnet 5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
Claude Sonnet 5.5 (Anthropic · proprietary)✓ verfügbar · 1M Kontext–
DeepSeek-V3.2 (DeepSeek · open weights)✓ verfügbar · 164k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
DeepSeek-V4.1-Flash (DeepSeek · open weights)✓ verfügbar · 1M Kontext (Preview)–
Gemini 3.1 Flash-Lite (Google · proprietary)✓ verfügbar · 1M Kontext–
Gemini 3.1 Pro (Google · proprietary)✓ verfügbar · 1M Kontext (Preview)–
Gemini 3.5 Flash (Google · proprietary)✓ verfügbar · 1M Kontext–
Gemini 3.5 Flash-Lite (Google · proprietary)✓ verfügbar · 1M Kontext–
Gemini 3.6 Flash (Google · proprietary)✓ verfügbar · 1M Kontext (EOL 19.11.2026)–
Gemini 3.7 Flash (Google · proprietary)✓ verfügbar · 1M Kontext (EOL 28.01.2027)–
Gemini 3.8 Flash (Google · proprietary)✓ verfügbar · 1M Kontext–
Gemini 4 Argon (Google · proprietary)✓ verfügbar · 1M Kontext–
Gemma 4 26B A4B (Google · open weights)✓ verfügbar · 262k Kontext (experimentell)–
Llama 3.3 70B Instruct (Meta · open weights)✓ verfügbar · 128k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
Llama 4 Maverick 17B 128E Instruct (Meta · open weights)✓ verfügbar · 524k Kontext–
Llama 4 Scout 17B 16E Instruct (Meta · open weights)✓ verfügbar · 1.3M Kontext–
Muse Spark 1.3 (Meta · proprietary)✓ verfügbar · 1M Kontext (Preview)–
Codestral 2 (Mistral AI · proprietary)✓ verfügbar · 128k Kontext–
Mistral Medium 3 (Mistral AI · proprietary)✓ verfügbar · 128k Kontext–
Mistral Small 3.1 (24B) (Mistral AI · open weights)✓ verfügbar · 128k Kontext–
Kimi K2 Thinking (Moonshot AI · open weights)✓ verfügbar · 262k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
gpt-oss-20b (OpenAI · open weights)✓ verfügbar · 131k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
gpt-oss-120b (OpenAI · open weights)✓ verfügbar · 131k Kontext–
Grok 4.3 (xAI · proprietary)✓ verfügbar · 200k Kontext (Preview)–
Grok 4.6 (xAI · proprietary)✓ verfügbar · 524k Kontext–
Grok 4.7 (xAI · proprietary)✓ verfügbar · 524k Kontext–
Grok 4.20 (xAI · proprietary)✓ verfügbar · 2M Kontext–
GLM-4.7 (Z.ai · open weights)✓ verfügbar · 200k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
GLM-5 (Z.ai · open weights)✓ verfügbar · 200k Kontext (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
GLM-5.2 (Z.ai · open weights)✓ verfügbar · 1M Kontext (MaaS endpoint guaranteed only until at least 30.11.2026) (Preview)–
Muse Ember 1.0 (Meta · proprietary)✓ verfügbar · 1M Kontext (Preview)–
Muse Wonder 1.0 (Meta · proprietary)✓ verfügbar · 1M Kontext (Preview)–
GPT-6 Luna (OpenAI · proprietary)✓ verfügbar · 1.1M Kontext (Private Preview; max. 922k input tokens) (Preview)–
GPT-6 Astra (OpenAI · proprietary)✓ verfügbar · 1.1M Kontext (Private Preview; max. 922k input tokens) (Preview)–
GPT-6.1 Terra (OpenAI · proprietary)✓ verfügbar · 1.1M Kontext (Private Preview; max. 922k input tokens) (Preview)–
GPT-6.1 Sol (OpenAI · proprietary)✓ verfügbar · 1.1M Kontext (Private Preview; max. 922k input tokens) (Preview)–
Gemini 4 Carbon (Google · proprietary)✓ verfügbar · 1M Kontext (access limited to verified enterprise customers)–
Claude Fable 5.2 (Anthropic · proprietary)✓ verfügbar · 1M Kontext (retirement not before 09.11.2027)–

Embedding Models

ServiceGoogle CloudHikube
Gemini Embedding 2 (Google)✓ verfügbar (multimodal (text, image, audio, video, PDF), 3072 dim, 8k tokens)–
gemini-embedding-001 (Google)✓ verfügbar (text, 3072 dim, adjustable output dimensionality)–
text-embedding-005 (Google)✓ verfügbar (legacy text embeddings, 768 dim)–
text-multilingual-embedding-002 (Google)✓ verfügbar (legacy text embeddings, 768 dim)–
multilingual-e5-large (Microsoft)✓ verfügbar (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–
multilingual-e5-small (Microsoft)✓ verfügbar (EOL 21.10.2026 (MaaS endpoint retired; self-deploy via Model Garden)) (auslaufend)–

Quantum Computing

ServiceGoogle CloudHikube
EmulatorCirq–
EngineQuantum AI–

Compliance & Certificates

ServiceGoogle CloudHikube
C5\nGerman cloud security catalogue✓ verfügbar–
ISO 27001\nInformation security management system✓ verfügbar✓ verfügbar
ISO 27017\nCloud security controls for providers✓ verfügbar–
ISO 27018\nPersonal data protection for cloud✓ verfügbar–
ISO 27701\nPrivacy information management extension✓ verfügbar–
ISO 50001\nEnergy management for data centers✓ verfügbar–
CSA STAR\nCloud Security Alliance assurance registry✓ verfügbar–
SOC 1,2,3\nService organization control audit reports✓ verfügbar–
PCI DSS\nPayment card data security standard✓ verfügbar–
ISO 20000\nIT service management system standard✓ verfügbar–
ISO 9001\nQuality management system standard✓ verfügbar–

Cost Indication

ServiceGoogle CloudHikube
Rating5.9/106.4/10
CPU vCPU/h3.9/10 0,0350278 €6.2/10 0,0119212 €
GPU H100/h3.6/10 9,5126750 €9.7/10 2,9785142 €
RAM GB/h5.3/10 0,0046939 €3.0/10 0,0119212 €
VMware CPU/h4.5/10 0,2544576 €–
Managed MySQL Ratio1/44.4/10 0,0715520 €4.2/10 0,0744966 €
Block Storage GB/h | 3000 IOPS7.5/10 0,0001107 €5.7/10 0,0002719 €
S3 Storage GB/h4.4/10 0,0000271 €4.0/10 0,0000298 €
Archive Storage GB/h6.8/10 0,0000071 €–
Tape GB/h7.1/10 0,0000029 €–
Kubernetes-Service-Premium Cluster/h5.3/10 0,0860000 €8.5/10 0,0476848 €
Serverless-Request per 1 Mil Requests4.4/10 0,3400000 €–
Serverless-Compute GB-Sec5.4/10 0,0000196 €–
Generative AI-Token Input gpt-oss-120b per 1Mil tokens10.0/10 0,0774000 €–
Generative AI-Token Output gpt-oss-120b per 1Mil tokens10.0/10 0,3096000 €–
Egress per GB3.0/10 0,1032000 €10.0/10 0,0000000 €

Einzelprofile