Add support for Deepseek V4 Pro and Deepseek V4 Flash

[HH45137]

Related GitHub Issue

Closes: #12173
Closes: #12177
Closes: #12174

Description

Added support for DeepSeek V4 Flash (deepseek-v4-flash) and DeepSeek V4 Pro (deepseek-v4-pro) models to the DeepSeek provider, as requested in #12173, #12177, #12174.

Changes made:

1. packages/types/src/providers/deepseek.ts — Added two new model entries to deepSeekModels:

   • deepseek-v4-flash
   • deepseek-v4-pro

2. src/api/providers/deepseek.ts — Updated the thinking model detection logic (line 59-62) to recognize the new V4 model IDs (deepseek-v4-pro and deepseek-v4-flash) in addition to deepseek-reasoner, enabling thinking mode for these models.

Key design decisions:

• V4 models use the same { thinking: { type: "enabled" } }

Test Procedure

• Open Roo Code in VS Code
• Select DeepSeek as API provider
• Verify deepseek-v4-flash and deepseek-v4-pro appear in the model dropdown
• Send a test message with each model to confirm they work
• Test thinking mode by sending a complex reasoning query
• Test tool calls with thinking mode to confirm reasoning_content is properly handled

Pre-Submission Checklist

• Issue Linked: This PR is linked to an approved GitHub Issue #12173
• Scope: Focused on adding DeepSeek V4 model support (one feature per PR)
• Self-Review: I have performed a thorough self-review of my code
• Testing: New and/or updated tests have been added to cover my changes (if applicable)
• Documentation Impact: No documentation updates required — models will appear automatically in the provider dropdown
• Contribution Guidelines: I have read and agree to the Contributor Guidelines

Screenshots / Videos

Documentation Updates

• No documentation updates are required.

Additional Notes

Pricing data for V4 models referenced from DeepSeek API Pricing and Thinking Mode Guide.

---

Interactively review PR in Roo Code Cloud

[hunterschuler]

I have tested this PR and it resolves #12173 and part of #12174 (regarding model entries) but does not resolve the other part of #12174 (400 errors) and #12177. Did this fix the 400 errors for you @HH45137?

[Aman9804]

I was using Roo code with deepseek-chat today and deepsheek dashboard reported I was using deepseek-v4-flash. Therefore I don't think this PR is needed anymore.

[MrGeennius]

deepseek-reasoner is currently using deepseek-v4-flash credits for me, even thought the model itself states that he is deepseek-v4-pro. It is for sure deepseek-v4-flash with the speed it has.

[HH45137]

I have tested this PR and it resolves #12173 and part of #12174 (regarding model entries) but does not resolve the other part of #12174 (400 errors) and #12177. Did this fix the 400 errors for you @HH45137?

On my end, Pro and Flash are functioning properly. I have used them to solve my C++ coding issues.