Currently, DeepSeek-R1's reasoning process is handled differently depending on the API provider:
## Azure Foundry & Fireworks API:
  • Model responses include <think> tags in the main response
  • Need to parse and separate these tags into a dedicated 'Thinking...' block
## OpenRouter:
  • <think> tags are not included in responses
  • Need an alternative method to capture and display reasoning tokens in the 'Thinking...' block
## DeekSeek Direct API:
#Feature request
Implement proper parsing and display of DeepSeek-R1's reasoning process in a dedicated 'Thinking...' block, with provider-specific handling:
  • Parse <think> tags from Azure Foundry and Fireworks API responses
  • Develop alternative method to capture reasoning from OpenRouter responses
This would create a consistent user experience for viewing model reasoning across all providers, similar to how DeepSeek's reasoning is displayed.
Suggested Implementation Priority:
  • Azure Foundry: Highest priority due to free API access, which would encourage more users to try DeepSeek-R1 through TypingMind
  • Fireworks API: Second priority due to superior token throughput, providing better user experience
  • OpenRouter: Lower priority due to slower token generation speed and more complex implementation requirements.