DeepSeek V4-Pro hands-on: Is max reasoning worse than high? Can Harness save it?
karminski3 · x · 2026-08-16
Blogger karminski3 releases a hands-on video of DeepSeek-V4-Pro-0813, testing four questions: whether the model is disappointing, whether max reasoning is worse than high or even v4-flash, and whether using DeepSeek Harness improves results. The video provides answers.
Related event: DeepSeek V4-Pro Tests Reveal Reasoning and Control Issues(2 posts)→
More from Models
- Visual Basis Atlas: 88 vectors control Grok Imagine's nuances — tetsuoai · 2026-08-16
- OpenAI's tokenizer is ~30% more efficient than Anthropic's — ns123abc · 2026-08-16
- Does LTX 2.5 still have a use case? Seemingly outperformed by Minimax — sdnr8 · 2026-08-16
- Netlify integrates OpenRouter and compares 11 AI models — philnash · 2026-08-16
- Grok Build reveals thought process, jokingly dubbed 'Open xAI' — burkov · 2026-08-16
- DeepSeek is a powerful model but requires careful output verification — Aizkmusic · 2026-08-16