Perplexity is retiring all of their current model lineup: The entire Llama-3.1-Sonar series stops working on Feb 22, 2025, replaced by the new Sonar lineup.

What you need to know:

• Legacy models will stop responding after Feb 22, 2025
• New models offer improved performance + simplified selection
• Sonar Pro brings expanded 200k context window

 Docs 

Anthropic is sunsetting all Claude 2.x series models + select Claude 3.0 models this July. This means, a clean migration required! These endpoints will stop responding completely.

Quick action items:

✅ Audit your Anthropic API calls

✅ Plan migration to Claude 3.5 Sonnet

✅ Finish all tests before July hits

Need help planning your migration? Drop a message on the forum here.

We've shipped several crucial updates to enhance cross-provider compatibility and reliability:

• Anthropic Enhancement: Added OpenAI-compatible total_tokens counter and improved stream error handling
• Gemini Tools Update: Smarter JSON schema handling for seamless OpenAI/Google model switching
• Pangea integration updated to latest API
• ContainCode plugin now has better Python and JS code block detection

Level up your LLM testing with Promptfoo support! Evaluate and assess your LLM outputs systematically through Portkey's Gateway.

Test and evaluate your LLM applications more efficiently with comprehensive evals on Promptfoo.

Connect your vector database workflows through Portkey!

We've added support for Milvus DB as a proxy router, enabling efficient vector data operations right from your Gateway.

Fine-tune your Gemini outputs with our new strictOpenAiCompliance flag:

Set true: Get clean, final responses with omitted CoT messages

Set false: Gateway will return both messages separate with \r\n\r\n separator

Mutator hooks are now live! Transform your request and response bodies with our new hooks, supporting key redaction plugins:

• portkey.redactPii
• pangea.redactPii
• patronus.redactPii
• patronus.redactPhi
• promptfoo.redactPii

We've streamlined how Gateway handles batches and files routes across Bedrock, Cohere, and OpenAI with our new unified format support. This means:

One consistent interface for all providers

• Simplified integration workflow
• Seamless provider switching
• Standardized batch processing

No need to handle different formats for different providers - Gateway takes care of it all for you!

✨How are companies actually deploying LLMs in production?

We went beyond the hype and analyzed 2T+ tokens to understand the real patterns when deploying LLMs in production - from provider dynamics to performance metrics.

 Link to Report 

Join us today as we break down insights from 2T+ LLM tokens across 90+ regions.

We analyzed real-world production data through Portkey's AI Gateway to understand how companies are actually implementing LLMs at scale.

Today we're sharing key insights around:

📊 Provider adoption trends in 2025

✨ Real-world performance metrics & benchmarks

⚡ Production patterns & cost optimization

Join us for the LLMs in Prod'25 launch webinar today at 10:00 AM PT :