Model deprecations

Stay informed about model lifecycle, deprecation timelines, and migration paths

Heroku Managed Inference and Agents regularly updates its model offerings to provide access to the latest and most capable AI models. This page outlines our deprecation policy and current model status.

Deprecation Policy

Timeline

When a model is scheduled for deprecation, we provide:

• 6 months notice - Announcement of deprecation date
• 3 months - Recommended migration window
• Final 3 months - Grace period before shutdown
• Shutdown date - Model becomes unavailable

Communication

We notify you through:

1. Dashboard notifications - Alerts in your Heroku Dashboard
2. Email notifications - Sent to account administrators
3. API warnings - Deprecation headers in API responses
4. Documentation updates - This page and release notes

Migration Support

During the deprecation period:

• Deprecated models continue to work normally
• Migration guides are provided
• New model recommendations are specified
• API compatibility is maintained when possible

Current Model Status

Active Models

Chat Models

Model	Status	Released
Claude Opus 4.5	✓ Active	2025
Claude 4.5 Sonnet	✓ Active	2025
Claude 4 Sonnet	✓ Active	2025
Claude 4.5 Haiku	✓ Active	2025
Amazon Nova 2 Lite	✓ Active	2025
Amazon Nova Pro	✓ Active	2024
Amazon Nova Lite	✓ Active	2024
Kimi K2 Thinking	✓ Active	2025
MiniMax M2	✓ Active	2025
Qwen3 235B	✓ Active	2025
Qwen3 Coder 480B	✓ Active	2025
OpenAI GPT OSS 120B	✓ Active	2024

Embedding Models

Model	Status	Released
Cohere Embed Multilingual	✓ Active	2024

Rerank Models

Model	Status	Released
Cohere Rerank 3.5	✓ Active	2024
Amazon Rerank 1.0	✓ Active	2024

Image Models

Model	Status	Released
Stable Image Ultra	✓ Active	2024

Deprecated Models

The following models are deprecated and scheduled for removal on February 28, 2025:

Action required: Migrate to the recommended replacement models before February 28, 2025 to avoid service interruption.

Model	Status	Removal Date	Recommended Replacement
Claude 3.7 Sonnet	🚨 Deprecated	Feb 28, 2025	Claude 4.5 Sonnet
Claude 3.5 Sonnet	🚨 Deprecated	Feb 28, 2025	Claude 4.5 Sonnet
Claude 3.5 Haiku	🚨 Deprecated	Feb 28, 2025	Claude 4.5 Haiku
Claude 3.0 Haiku	🚨 Deprecated	Feb 28, 2025	Claude 4.5 Haiku

These models continue to function normally until the removal date. We recommend migrating as soon as possible to benefit from improved capabilities in the newer models.

Deprecation Status Indicators

Models go through several stages before deprecation:

✓ Active

• Fully supported and recommended
• Receives updates and improvements
• No deprecation planned

⚠️ Maintenance Mode

• Still supported but not receiving updates
• Consider migrating to newer models
• Deprecation may be announced

🚨 Deprecated

• Scheduled for removal
• Migration recommended immediately
• Specific end-of-life date announced

❌ End of Life

• No longer available
• API requests will fail
• Must use alternative models

How to Check Model Status

Via Dashboard

1. Log into Heroku Dashboard
2. Navigate to your app
3. Check for deprecation notices in notifications
4. Review model resource details

Via API Headers

Deprecated models return warning headers:

curl -I https://us.inference.heroku.com/v1/chat/completions \
  -H "Authorization: Bearer $INFERENCE_KEY"

# Response includes:
# Warning: 299 - "This model is deprecated and will be removed on 2025-XX-XX"
# Sunset: Sat, 01 Jun 2025 00:00:00 GMT

Via CLI

# Check your model status
heroku ai:models:info --app my-app

# List available models (shows deprecation status)
heroku ai:models:list

Migration Planning

When You Receive a Deprecation Notice

Follow these steps to ensure smooth migration:

1. Review Timeline

• Note the deprecation date
• Plan migration before the deadline
• Allow time for testing

2. Identify Recommended Replacement

We provide specific migration paths for each deprecated model:

Example:
Old Model: claude-3-5-sonnet
Replacement: claude-4-5-sonnet

3. Test in Development

# Create a new model resource for testing
heroku ai:models:create claude-4-5-sonnet \
  --app my-app-staging \
  --as INFERENCE_NEW

# Test your application
# Compare outputs with old model
# Verify performance and quality

4. Update Application Code

If model IDs are hardcoded, update them:

# Before
model = "old-deprecated-model"

# After
model = "claude-4-5-sonnet"

Or use environment variables:

# Recommended approach
model = os.getenv("INFERENCE_MODEL_ID")

5. Deploy to Production

# Attach new model to production app
heroku ai:models:create claude-4-5-sonnet \
  --app my-app-production \
  --as INFERENCE

# Deploy updated code
git push heroku main

# Verify everything works
heroku logs --tail --app my-app-production

6. Clean Up

# After successful migration, remove old model
heroku ai:models:detach OLD_MODEL --app my-app-production

Backward Compatibility

API Compatibility

When possible, new models maintain API compatibility:

• Same request/response format
• Compatible parameter sets
• Similar behavior patterns

Breaking Changes

If a replacement model has breaking changes, we provide:

• Detailed migration documentation
• Code examples showing differences
• Extended deprecation timeline

Example Migration Scenarios

Scenario 1: Simple Model Upgrade

Situation: Claude 3.0 Opus deprecated → Claude 4 Sonnet recommended

Migration Steps:

1. Test new model

   heroku ai:models:create claude-4-sonnet --app my-app-staging
   
2. Update environment variable

   heroku config:set INFERENCE_MODEL_ID=claude-4-sonnet --app my-app
   
3. Verify

   Test your application to ensure it works correctly

4. Remove old model

   heroku ai:models:detach INFERENCE_OLD --app my-app
   

Scenario 2: Feature Change Required

Situation: Model deprecation requires code changes

Migration Steps:

1. Review migration guide

   Read the specific migration documentation for your model

2. Update code

   Implement required changes in your codebase

3. Test thoroughly

   # Create new model
   heroku ai:models:create new-model --app my-app-staging
   
   # Run your test suite
   pytest tests/
   
4. Deploy

   # Deploy to staging first
   git push staging main
   
   # Then to production after verification
   git push heroku main
   

Best Practices

Proactive Monitoring

Subscribe to Updates

• Enable email notifications in your Heroku account
• Check the dashboard regularly for notices
• Monitor API response headers
• Review this page periodically

Use Environment Variables

# Don't hardcode model names
# Bad:
model = "claude-3-sonnet"

# Good:
model = os.getenv("INFERENCE_MODEL_ID")

This makes model switching much easier during migrations.

Test New Models Early

Don’t wait until the last minute:

• Test recommended replacements as soon as deprecation is announced
• Compare quality and performance
• Identify any integration issues early

Plan for Multiple Environments

Maintain separate model resources for:

• Development
• Staging
• Production

This allows safe testing before production migration.

Frequently Asked Questions

What happens if I don't migrate before the deadline?

After the end-of-life date, the deprecated model will stop working:

• API requests will return errors
• Your application will fail
• You’ll need to migrate immediately

Avoid this: Migrate well before the deadline with proper testing.

Can I still use a deprecated model?

Yes, during the deprecation period (usually 6 months), deprecated models continue to work normally. However, we strongly recommend migrating as soon as possible to:

• Avoid last-minute issues
• Access improved capabilities
• Take advantage of better pricing

Will pricing change during deprecation?

Pricing for deprecated models typically remains the same during the deprecation period. However, new models may have different pricing structures. Check the pricing page for details.

How do I know which model to migrate to?

We provide specific migration recommendations for each deprecated model:

• Check the deprecation notice in your dashboard
• Review this page for recommended replacements
• See the choosing a model guide
• Test alternatives in AI Studio

What if the replacement model doesn't work for my use case?

If the recommended replacement isn’t suitable:

1. Review all available models on the models overview page
2. Test alternatives in AI Studio
3. Contact Heroku support for guidance
4. Consider adjusting your application to work with available models

Will my data be affected?

No. Model deprecations only affect:

• Which models are available for inference
• API endpoints for those models

Your application data, configurations, and other resources remain unchanged.

Getting Help

If you have questions about model deprecations or need migration assistance:

1. Documentation: Review the models overview and choosing a model guides
2. AI Studio: Test replacement models interactively
3. Heroku Support: Contact support through your dashboard
4. Community: Ask questions in Heroku forums

Stay Informed

To stay up-to-date on model changes:

• Enable notifications in your Heroku Dashboard
• Check this page regularly
• Monitor the release notes
• Subscribe to Heroku AI announcements

Proactive migration is key: Don’t wait until the last minute. Test and migrate as soon as deprecations are announced to ensure smooth transitions.

Related Resources

Models overview

See all available models

Choosing a model

Select the right model for your needs

Pricing

Compare model costs

CLI Commands

Manage model resources