GPU
Providers using this tag (93)
Ranked by API Evangelist rating — Exemplar and Strong are expanded by default.
Exemplar 2 Complete, well-documented, and agent-ready
Strong 11 Solid coverage with minor gaps
Developing 20 Usable, with meaningful gaps to close
Thin 25 Limited public surface area
Emerging 20 Early or largely undocumented
Minimal 14 Almost no public developer surface
Unrated 1 Not yet scored
APIs with this tag (44)
Ranked by the provider's API Evangelist rating — the Kin Score is scored per provider, not per API, so every API of a provider shares its band. How the rating works →
Exemplar 3 Complete, well-documented, and agent-ready
Strong 6 Solid coverage with minor gaps
fal Compute APIProvision and manage dedicated GPU instances (H100, H200, A100, B200) with full SSH access for training, fi...Paperspace Deployments APIContainer-as-a-service deployments that run user-provided images on Paperspace GPU machines with a managed ...Paperspace Machines APIProgrammatically manage Paperspace virtual machines — GPU and CPU compute instances. Covers the machine lif...Modal Images and GPUDefine container images programmatically with modal.Image (from_registry, debian_slim, pip_install, run_com...TensorDock Instances APIThe current TensorDock Instances API (v2) at https://dashboard.tensordock.com/api/v2/instances for creating...TensorDock Secrets APIThe TensorDock Secrets API (v2) for managing SSH keys and generic secrets that are encrypted at rest and in...
Developing 12 Usable, with meaningful gaps to close
Hyperbolic GPU Marketplace APIDecentralized on-demand GPU compute marketplace renting idle H100, H200, A100, and RTX 4090 capacity from t...Hyperbolic On-Demand GPU APIBare-metal GPU rental with SSH access for full control. Hourly rates from $1.39-$1.99 with no long-term com...Lambda 1-Click ClustersLambda 1-Click Clusters provision interconnected clusters of 16 to 2,000+ NVIDIA H100 SXM or B200 GPUs for ...Lambda Cloud APIThe Lambda Cloud API is the REST control plane for launching, listing, starting, stopping, and terminating ...RunPod GraphQL APIThe RunPod GraphQL API provides programmatic access to Pods, templates, and Serverless endpoints via GraphQ...RunPod ServerlessRunPod Serverless provides pay-as-you-go inference endpoints with autoscaling workers, queue-based and load...Vast.ai REST APIThe Vast.ai REST API is the control plane for the GPU marketplace. It supports searching offers across host...Akash Console Network Data APIPublicly accessible indexed API for querying provider information, GPU availability, resource statistics, a...Akash Network Gpu APIThe Gpu API from Akash Network — 4 operation(s) for gpu.Novita AI GPU APIOn-demand GPU instance management and templates plus serverless GPU endpoints. Create, start, stop, and del...LanceDB EnterpriseDistributed, multi-tenant multimodal lakehouse. Adds curation and deduplication, Python UDF feature enginee...Colab Runtime and Kernel ManagementGoogle Colab provides internal APIs for managing notebook runtimes and kernels, including connecting to hos...
Thin 12 Limited public surface area
CoreWeave Kubernetes Service APIThe CKS API provisions and manages CoreWeave Kubernetes Service clusters and node pools on bare-metal GPU a...AnyscaleAnyscale is the production-scale AI platform built on Ray by the creators of Ray, supporting LLM inference ...NVIDIA NIMNVIDIA NIM is a set of inference microservices for streamlined AI model deployment, prebuilt and optimized ...Together AITogether AI is a full-stack AI Native Cloud for inference, fine-tuning, and GPU clusters powered by researc...IO Cloud Container-as-a-Service APIContainer-as-a-Service REST API for deploying GPU-backed containers across io.net's decentralized network. ...IO Cloud MCP ServerHosted Model Context Protocol server at https://mcp.io.solutions/mcp that exposes IO Cloud provisioning and...IO Cloud VM-as-a-Service APIVM-as-a-Service REST API for provisioning, extending, and terminating GPU virtual machines on io.net. Surfa...Shadow GPU (OpenStack API)Shadow GPU is Shadow's sovereign GPU cloud, delivered as a curated OpenStack 2024.1 deployment rather than ...NVIDIA Triton Inference Server HTTP APINVIDIA Triton Inference Server is an open-source inference serving software that implements the KServe Open...vLLM OpenAI-Compatible APIvLLM is a high-throughput and memory-efficient inference engine for LLMs, implementing PagedAttention for e...PennyLane LightningThe Lightning plugin family — `lightning.qubit` (C++ CPU), `lightning.gpu` (cuQuantum), and `lightning.kokk...Cumulus Inference GatewayOpenAI-compatible HTTP inference gateway. One client works against every upstream provider; per-workflow ro...
Emerging 8 Early or largely undocumented
Nebius Compute APIThe Nebius Compute API provisions and manages virtual machines and GPU clusters with NVIDIA GPUs and Infini...Nebius Managed Kubernetes APIThe Managed Kubernetes API provisions Kubernetes clusters with GPU and InfiniBand support for distributed t...HorovodHorovod is a distributed deep learning training framework for TensorFlow, Keras, PyTorch, and Apache MXNet.Newton Physics EngineNewton is a Python library and physics simulation engine. It is consumed programmatically via Python (not a...Spheron GPU Cloud APIREST API for programmatically managing GPU deployments across the Spheron marketplace. Covers provider and ...Pasqal Emulators (EMU-MPS, EMU-SV)Pasqal's PyTorch-based GPU-accelerated emulator monorepo housing EMU-MPS (matrix-product-state emulator for...Unreal Render Hardware Interface (RHI)The Render Hardware Interface is Unreal's abstraction over graphics APIs (Direct3D 11/12, Vulkan, Metal, Op...Spector.jsBrowser extension and embeddable library that captures and inspects WebGL / WebGPU frames — every draw call...
Minimal 3 Almost no public developer surface
NVIDIA DGX Cloud Lepton APIManaged AI platform API exposing model endpoints (HTTP server with OpenAI-compatible chat surface), dev ses...NVIDIA Developer APINVIDIA provides developer APIs for GPU computing, AI/ML model training and inference, and graphics. The pla...JD Cloud Compute (VM) APIJD Cloud Compute APIs manage Cloud Virtual Machines (CVM), lightweight VMs, GPU instances, bare metal serve...
Companies reaching this through an API (11)
These companies publish an API, specification or operation carrying “GPU” but do not classify their business under it. Listed unranked and kept out of the count above, because one tagged operation is not a statement about what a company does.
AI Gateway
Babylon.js
Google Colab
JD.com (Jingdong)
LanceDB
LF AI and Data
Pasqal
Scalable Inference Serving
Scope3
Unreal Engine
Xanadu
Score breakdown
Frequency
57.8
log-scaled weighted occurrences
Breadth
2.1
spread across providers
Quality lift
34.5
mean composite of providers using it
Cohesion
21.1
strength of nearest seed neighbor
Related tags
Inference 47 co-occurrences
Compute 35 co-occurrences
OpenAI-Compatible 20 co-occurrences
AI Infrastructure 16 co-occurrences
Cloud Computing 13 co-occurrences
Chat Completions 16 co-occurrences
Where this tag comes from
Provider tag97
Api tag45
Openapi tag3
Openapi op tag10
Markets provisional
Kin Score markets whose measured providers carry this tag. Share is the fraction of the market's measured providers that carry it — it describes the market, not this tag.
| Market | Share of the market's providers |
|---|---|
| MLOps and AI/ML Infrastructure (LF AI & Data, MLPerf, Kubeflow) | 21% |
Cohort brief
Auto-generatedThe 90 providers in the APIs.io catalog tagged GPU, scored on the Kin Score. Every figure below is computed from the catalog — nothing here is written.
Providers
90
of 93 in the catalog
Mean Kin Score
33.3
+11.1 vs catalog 22.2
Mean Agent Readiness
17.9
+5.9 vs catalog 12
Spread
0–66.9
median 34.6 · σ 17.8
How the 90 split by band
Exemplar 2Strong 11Developing 20Thin 25Emerging 19Minimal 13
Facet averages, against the whole catalog
| Facet | This cohort | Catalog | Difference | Scored |
|---|---|---|---|---|
| Developer Ergonomics | 42.1 | 23.2 | +18.9 | 90 |
| Contract Quality | 29.6 | 17.2 | +12.4 | 90 |
| Access Clarity | 36.3 | 23.9 | +12.4 | 90 |
| Discoverability | 64.6 | 57.9 | +6.7 | 90 |
| Operational Transparency | 19.3 | 13 | +6.3 | 90 |
| Contract Governance | 8.3 | 6.5 | +1.8 | 90 |
A facet is averaged over the members that carry it, not over the whole cohort — the “Scored” column is that count. Averaging an absent facet as zero would score our own coverage gaps as the providers’ posture.
What this cohort publishes
| Artifact | This cohort | Catalog | Difference |
|---|---|---|---|
| MCP server (any) | 22% | 13% | +9 |
| MCP server (first-party) | 22% | 12% | +10 |
| Agent Skills | 0% | 0% | 0 |
| OAuth scopes | 4% | 11% | -7 |
| Security | 97% | 98% | -1 |
| Arazzo workflows | 4% | 2% | +2 |
| Governance rules | 23% | 16% | +7 |
mcp_pct counts any mcp/ artifact including ones API Evangelist derived from the provider OpenAPI; mcp_first_party_pct counts only servers the provider publishes. Prefer the latter.
Top by Kin Score
- 1 Segmind 66.9
- 2 NexGen Cloud 66.5
- 3 fal 65.6
- 4 Paperspace 64.2
- 5 SAKURA internet 63.7
- 6 NVIDIA NIM 60.6
- 7 Mithril 60.5
- 8 RightNow AI 59.5
- 9 The San Francisco Compute Company 57.2
- 10 Simplismart 57
Top by Agent Readiness
- 1 The San Francisco Compute Company 55.9
- 2 Vultr 49.8
- 3 fal 48.1
- 4 NexGen Cloud 43.3
- 5 Thunder Compute 42.4
- 6 Polar Signals 41.2
- 7 OpenRelay 40.4
- 8 FriendliAI 39.6
- 9 Fireworks AI 39.4
- 10 Vast.ai 39
All 93 members of this roster resolve to a provider in the
catalog; 90 carry a score (97%).
The figures above describe the scored members only.
Coverage caveat. 74 of 90 scored members had under half their catalog-attributable points satisfied at all (catalog_earned / catalog_max < 0.5). They are INCLUDED in the mean above. A large share here means the cohort mean is partly a measurement of how far our enrichment reached, not only of how these providers publish.Generated from the catalog build of 24 October 2026, across 31984 providers, using the same computation served by the APIs.io cohort API. Briefs are published for rosters of 5 or more scored providers; below that a distribution is not meaningful.
Coverage caveat. 74 of 90 scored members had under half their catalog-attributable points satisfied at all (catalog_earned / catalog_max < 0.5). They are INCLUDED in the mean above. A large share here means the cohort mean is partly a measurement of how far our enrichment reached, not only of how these providers publish.Generated from the catalog build of 24 October 2026, across 31984 providers, using the same computation served by the APIs.io cohort API. Briefs are published for rosters of 5 or more scored providers; below that a distribution is not meaningful.
Work with this as data
Every tag here is available over the APIs.io API and to AI agents over MCP.