DeepSeek V4's Reasoning Style Becomes More Like Claude
teortaxesTex · x · 2026-07-18
The shared post discusses a new reasoning example from DeepSeek V4: the model's output style now resembles Claude more closely.
Specific changes include:
- No longer outputting massive walls of code within the reasoning chain
- Shifting to more concise, step-by-step descriptions like "writing main scene code" or "writing core functions"
- Providing more summaries in responses, making the overall output clearer and more readable
The reposter expressed a notably negative attitude towards these changes, feeling that "things are moving in a direction I don't like."
More from Models
- Claim says Kimi was distilled from Fable, sparking a model-attribution jab — cephaloform · 2026-07-22
- Gary Marcus says LLM math skills are like knowing only a car’s engine size — GaryMarcus · 2026-07-22
- OpenAI’s Codex + GPT-5.6 Sol hits 99% recall in Project APE verification tests — soumitrashukla9 · 2026-07-22
- OpenAI-linked paper says capability RL can make models more reward-seeking — MariusHobbhahn · 2026-07-22
- Macaron V1 adds LoRA RL on GLM 5.2 and claims SOTA benchmark gains — Xianbao_QIAN · 2026-07-22
- OpenAI rolls out voice in GPT-Live, but the UI obscures search and reasoning — Graham_dePenros · 2026-07-22