/
Select a project
New Foundry
ST

Welcome to Microsoft Foundry

Pick a project to start building — deploy models, create agents, and monitor usage. Each project is scoped to one Azure region.

New here? You’re in the right place.
Open an existing project or create one to get API keys, browse 118+ models, and ship your first agent in minutes.
Quickstart
  1. 1Select a project above — check region and health before you open it.
  2. 2Grab credentials from Home, then deploy a model from Discover.
  3. 3Test in Playground, create an agent, and monitor usage in Operate.
Help & community
Tips
Projects share one region quota. Use Discover → Models to compare latency and cost before deploying.

Welcome, Swaraj Thamke

test-1 · East US 2
Project credentials
API key
sk-az-••••••••••••••••2f9a
Project endpoint
https://test-1-resource-swaraj.services.ai.azure.com
Azure OpenAI endpoint
https://test-1-resource-swaraj.openai.azure.com
2
Active agents
5
Deployments running
118
Models available
East US 2
Project region
Start here
Continue where you left off View all →
genie Agent v6
Modified 2 hours ago · prompt
Running
testeagent Agent v1
Modified yesterday · prompt
Running
gpt-chat-latest-2 Deployment
Deployed 5/29/26 · Global Standard
Succeeded
claude-opus-4-8 Model
Viewed 3 days ago · Anthropic · Messages
New arrivals View all →
claude-opus-4.8 New
Anthropic · Messages
Most intelligent Opus model for coding, agents, and enterprise work.
gpt-realtime-2 New
OpenAI · Audio generation
Low-latency speech-to-speech with built-in reasoning.
DeepSeek-V4-Pro New
DeepSeek · Chat completion
Highly efficient intelligence with million-token context.

Discover what's possible

Explore frontier AI from the world's leading providers — all accessible from one place, secured by Azure.

118Models
30Agent starters
1,604Tools
20+Providers
Explore by provider
View all models →
Model leaderboard preview
Quickly compare quality, safety, speed, and cost before opening the full ranking.
Full leaderboard →
Featured models
View all 118 →
Model collections
Direct from Azure
Flagship models from leading providers, securely delivered with Azure support + seamless integration.
Foundry Labs
Built by Microsoft Foundry, in partnership with Microsoft Research.
Foundry Local ↗
Run models locally to reduce costs and keep data on your device.
New arrivals
See all new models →
Models by Microsoft Foundry
Phi, MAI, and Florence — co-developed models built for Azure deployment.
View all Microsoft models →
Model Leaderboard
Compare AI model performance across quality, safety, throughput, and cost. Updated May 2026.
Benchmark results reflect controlled evaluations and may not represent real-world performance for your specific use case. Scores are compared relative to each other—higher quality index and higher throughput indicate better performance, lower safety refusal rate and lower benchmark cost indicate better results in those dimensions.
Performance at a glance
All models
# Model Quality index i ↑ ↓ Safety refusal i ↑ ↓ Throughput (t/s) i ↑ ↓ Benchmarks i ↑ ↓ Benchmark cost i ↑ ↓
Quality vs cost trade-off Bubble size = throughput
Best zone: higher quality + lower cost
Quality index →
← Lower benchmark cost is better ($/M tokens)
Leaderboards by scenario Top 5 per category

Agents

Preview

Get started with agent development using starter manifests.

Starter manifests help you learn and experiment as you build your own agents, but they are for educational use only. Agents created from them aren't production ready, so review risks and ensure compliance with your organization's policies, laws, and safety requirements before use.

Models

Sort
All models
Available in my project
East US 2
All
Low (<$1/1M)
Medium ($1–5/1M)
High (>$5/1M)

Models

Name Model Version Deployment status Deployment type Created on Created by

Agents

Saved agent configurations ready to deploy or edit

No agents yet
Configure a model in the Playground and click Save as agent to see it here.

Overview

Preview
May 25, 2026 – Jun 1, 2026
Non-Foundry agent data is always displayed. Showing data for project test-1.
0 Active alerts
Running agents
2 / 2
All healthy
View all agents →
Estimated cost
$124
↑ 18% vs last period
Agent success rate
98.4%
↑ 0.3% vs last period
Token usage
4.2M
Tokens this period
Agent run volume Calls over time
Top increases: testeagent ↑42%
Top decreases: genie ↓8%
Agent success rate % per day
Top increases: genie ↑1.2%
All agents above 95%
Recent agent runs
Agent Status Tokens Duration Time
Foundry AI
Ask for models, providers, or use cases. I can jump the catalog to a useful filtered view.

Model Comparison

Deployment Info

Model
Provider
How do you want to deploy this model?
Model deployed successfully
Your model is now a live deployment with its own endpoint and API key.
Endpoint created
https://my-foundry.openai.azure.com/openai/deployments/gpt-4o
What would you like to do next?

Assets

Preview
Project: test-1
Name Source Project Status Version Error rate Est. cost Token usage Runs Monitoring