User Complaints: DeepSeek Ignores Rule Prompts, Falls Behind Qwen in Practice
Juulk9087 · reddit · 2026-08-02
A user running the DeepSeek model natively locally criticized the model for its inability to effectively follow user-defined rules and system prompts during actual coding tasks.
The user reported that the model ignores customized behavioral constraints regardless of using first or second-person prompts, or whether the prompts are in Chinese or English. They argue that this inability to be steered via rules is why many benchmark-heavy models fail in real-world, complex user environments. Frustrated by the model's subpar behavior, the author has decided to switch back to Qwen 27b.
More from Models
- Grok Voice 3.0 tops speech-to-speech index, undercuts OpenAI on price — VectorWireAI · 2026-10-16
- Microsoft completes MAI-Base-2, a 2.8T-parameter MoE trained on 16k GB200 GPUs — VectorWireAI · 2026-10-16
- Gemini Pro users hit sudden usage paywalls: 100% daily limit despite 6% weekly — Strange-Plate581 · 2026-10-16
- Users report Gemini's memories can't actually be deleted even after wiping chat history — ThrowRA_Purpose2025 · 2026-10-16
- OpenAI stealth-testing mysterious internal model 'Aeon' in Codex desktop app and Dots — cedric_chee · 2026-10-16
- Argon edges out astra as daily driver, but Claude still wins heavy coding work — iruletheworldmo · 2026-10-16