Skip to content
Sign inRequest access

Products

GPU CloudAMD MI325X and NVIDIA H200 by the hour or on reservation.
Dedicated ClustersPreviewPrivate, interconnected GPU clusters for sustained workloads.
SandboxesPreviewIsolated, short-lived environments for developers and AI agents.
Serverless GPUPreviewGPU containers that scale with requests, down to zero.
Serverless InferenceOpen models behind one OpenAI-compatible endpoint.
Dedicated EndpointsReserved serving capacity for one model, billed per GPU-hour.
PlaygroundSend a prompt to any model your workspace can reach.
Model DeploymentsPreviewImmutable deployments with instant promote and rollback.
Fine-tuningLoRA fine-tunes on the open models in our catalog.
Model TrainingPreviewFoundation-model training runs, managed end to end.
EvaluationsPreviewBenchmark suites and model comparisons you can rerun.
DatasetsManaged and synthetic training data, versioned and ready to use.
Data ManagementPreviewDatasets, connections and storage, with lineage across the platform.
PipelinesPreviewVersioned DAGs that automate data, training and deployment.
Data RoomsPreviewSecure document sharing with guests, NDAs and an audit trail.
NokuvaPreviewThe AI-native design editor.
TavocPreviewA Rust-native code platform: IDE and desktop apps.
FoltracComing soonMulti-cloud deployment with autonomous operations.
W0rkTreeComing soonVersion control for agent fleets.
Identity & AccessMembers, roles, policies, SSO and audit for every workspace.
Billing & CostUsage, invoices, credits and budgets in one place.
Logs & MonitoringRequest logs, latency and error rates across services.
Project CanvasPreviewSee and change a project's resources on one live canvas.

Pages

Home/
Platform/platform
All products/products
Pricing/pricing
GPU pricing/pricing#compute
Inference services/services/inference
Model training services/services/model-training
Protocol design/services/protocol-design
Research partnerships/services/research-partnerships
Protocols/protocols
Datasets/datasets
Documentationdocs.legionedge.ai
Changelog/changelog
News/news
Research/research
Blog/blog
Articles/articles
Events/events
Startup Program/programs/startup
Ambassador Program/programs/ambassador
Partner Program/programs/partner
Event Program/programs/events
Company/company
Careers/careers
Partners/partners
Contact/contact
Statusstatus.legionedge.ai
Terms of Service/legal/terms
Privacy Policy/legal/privacy

Serverless Inference

Available

Call open models with the code you already have.

Call DeepSeek, Qwen, GLM and Kimi models through inference.legionedge.ai/v1 with the client libraries you already use. Pay per token, with no capacity to manage.

Request accessView pricing
Pricing
From $0.10 / 1M tokens
Layer
Inference
Serving models from
DeepSeek, Qwen, GLM, Kimi, Your fine-tuned adapters

Why Serverless Inference

No capacity to plan, no client to rewrite.

Serverless inference runs catalog models on shared capacity that LegionEdge scales for you. Point the client libraries you already use at inference.legionedge.ai/v1, pick a model, and pay only for the tokens you send and receive.

  • OpenAI-compatible

    Point your SDK’s base URL at inference.legionedge.ai/v1 and keep your code. Requests and responses keep the shape your client already expects.

  • Pay per token

    Input and output tokens are priced per model and billed per request, with no capacity to reserve and nothing to keep warm.

  • Request logs

    Latency, tokens and errors for every call, searchable by model, status and key.

How it works

Open models behind one OpenAI-compatible endpoint.

  1. Pick a model

    Browse the catalog by provider, context window and price per token, and see which models you can also fine-tune.

  2. Create a key

    Issue an API key scoped to invoking inference, with request and token rate limits set per key.

  3. Point your client at v1

    Set your SDK’s base URL to inference.legionedge.ai/v1. The rest of your code stays as it is.

  4. Watch every call

    Request logs show latency, tokens and errors for each call, and billing shows what you spent by model and by day.

In the console

Everything in one place.

Serverless Inference lives in the LegionEdge cloud console, next to every other service in the workspace, under one account and one identity.

  • Model catalog

    Open models with their provider, context window and input and output prices, in one list.

    • DeepSeek, Qwen, GLM and Kimi families
    • Multimodal and prefix-caching models marked
    • Which models you can fine-tune

  • Managed projects

    Serve a catalog model, or your fine-tuned adapter on a shared base, as a project with its own URL.

    • Default model parameters per project
    • Rate limits and guardrails
    • No cold start on shared capacity

  • Keys and limits

    API keys from Identity & Access, scoped to the projects they may call.

    • Keys scoped to invoking inference
    • Request and token limits per key
    • Allowed keys and IP ranges per project

  • Logs and spend

    Every request is inspectable, and every token shows up on the bill it belongs to.

    • Filter logs by model, status, key and latency
    • Prompt and response logging off by default
    • Spend by service, model and day

Pricing

From $0.10 / 1M tokens

Serverless inference is billed per million input and output tokens, per request, with no capacity to reserve.

  • Input and output tokens are priced separately for each model.
  • Serverless inference and the Playground are on every plan, including Researcher.
  • Need capacity of your own? Dedicated endpoints are billed per GPU-hour.
Full pricing

Catalog models

Catalog models with context window and token prices
ModelContextInput / 1MOutput / 1M
DeepSeek V4 Flashdeepseek-v4-flash160K$0.30$1.20
DeepSeek V4 Prodeepseek-v4-pro800K$1.74$3.48
Qwen3 0.6Bqwen3-0.6b32K$0.10$0.20
Qwen3 4B Instructqwen3-4b-instruct32K$0.20$0.60
Qwen3.6 27Bqwen3-6-27b256K$0.30$0.90
Qwen3.8 27Bqwen3-8-27b256K$0.30$0.90
GLM-5.2glm-5-2128K$0.50$2.00
GLM-5.3glm-5-31M$0.50$2.00
Kimi K3kimi-k31M$0.60$2.50
Kimi K3 Turbokimi-k3-turbo128K$0.60$2.50

Related products

Pairs well with Serverless Inference

Serverless Inference

Keep your code. Change your model.

Point your client at v1 and start with any model in the catalog.

Request accessTalk to us