← Back to blogchangelog

Try GPT-4.1 now + more updates

Zef

Zef

~ 1 min read


Try GPT-4.1 now + more updates

Hi agent builders! It’s been a while but we’ve got some great updates for you today!

Introducing GPT-4.1 Models

Hot off the press, only released yesterday, we've integrated OpenAI's latest GPT-4.1 models into Chat Thing, offering significant performance enhancements across the board. You can't use these models via ChatGPT as they are API only so Chat Thing is a great place to test them out! Here's some of the improvements:

  • Expanded Context Window: Handle up to 1 million tokens, enabling deeper context understanding.​
  • Improved Coding Capabilities: Achieved a 54.6% score on SWE-bench Verified, surpassing GPT-4o's 33.2%. ​
  • Enhanced Instruction Following: Better at adhering to complex, multi-step instructions. ​
  • Cost-Effective Options: GPT-4.1 Nano is now our most affordable model.

Read OpenAI’s announcement post here

You can select the 4.1 models from the model selector in your bot settings now.

Enhanced Chat Responses

  • Code Syntax Highlighting: Code blocks in responses now feature syntax highlighting for improved readability.​ Code Block Syntax Highlighting PR 851.png
  • LaTeX Rendering: Mathematical expressions are rendered using LaTeX for clearer presentation.​ LaTeX Math Expression Rendering PR 852.png

Additional Updates

  • Multilingual Support: Added and refined translations for multi-modal chat interface elements.​
  • RSS Data Source Scraping: Enabled CSS exclude rules to refine content extraction.​
  • Email Enhancements: Included reply-to headers in "talk-to-human" emails for better communication.​
  • UI Improvements: Enhanced conversation view styling for better readability.​
  • Bug Fixes: Resolved issues including duplicate bot-to-bot responses and improved API power-up validation.​

We can’t wait to see what you think of these updates, happy agent building!

Related articles

Test blog post three

Test blog post three

Mauris consectetur nibh quis tempus viverra. Vivamus dapibus mi turpis, non ultrices nibh mollis et. Fusce vitae dui et nibh malesuada vestibulum.