2026-06-24

• Coding Agent: Claude Code / Kimi Code / OpenCode / MiMo Code
• Main Dev: GLM-5.2 / Kimi-K2.7 / MiniMax-M3
• Code Review: DeepSeek-V4-Flash
• Talk: Kimi-K2.6

2025-12-16

• Code Agent: Sonnet 4.5
• Talk: Opus 4.5

2025-10-24

• Code Agent: Kilo Code - GLM 4.6
• Code Agent: Claude Code - Sonnet 4.5 or Deepseek Reasoner (v3.2)
• Complex reasoning: Claude Sonnet 4.5
• Daily conversation: GPT-5

2025-10-09

Current workflow:

• Code Agent: Kilo Code - GLM 4.6
• Code Agent: Claude Code - GLM 4.6
• Complex reasoning: deepseek/deepseek-v3.2
• Hard problem solving: Claude Sonnet 4.5
• Daily conversation: GPT-5

2025-09-22

Current workflow:

• PLAN mode: Claude Code - Claude Sonnet 4
• ACT mode: deepseek/deepseek-v3.1 (for complex reasoning)
• Daily conversation: GPT-5

2025-08-08

Workflow:

• PLAN mode: deepseek/deepseek-reason (on https://platform.deepseek.com)
• ACT mode: qwen/qwen3-235b-a22b-07-25 (for quick code editing)

2025-07-27

Aliyun’s official qwen is too expensive, 50 yuan is gone. And based on leaderboard tests, performance is worse than r1-0528.

• Fast programming model: deepseek/deepseek-v3
• Planning model: deepseek/deepseek-r1-0528
• Daily conversation: chatgpt-4o

2025-07-26

• Fast programming model: qwen/qwen3-235b-a22b-07-25
• Planning model: deepseek/deepseek-r1-0528
• Daily conversation: chatgpt-4o

Previous

• Fast programming model: cursor-auto
• Planning model: deepseek/deepseek-r1-0528
• Daily conversation: chatgpt-4o