DeepSeek V4 Pro 0813 fails same demo twice while Flash succeeds, user reports
neverontime5 · reddit · 2026-08-14
Reddit user neverontime5 reports initial testing of DeepSeek V4 Pro 0813: the same demo failed twice on Pro, while V4 Flash completed it. Generation speed was 80-90 tokens/s, but the task wasn't finished. The user notes the small sample size and plans to use ZenMux for more detailed comparison.
More from Models
- Dev Quips: Literally No One Uses bf16 Anymore — GavinSBaker · 2026-08-14
- What's your favorite oddball model? Redditor recommends Nambeige 3B — Ecstatic-Wash-7667 · 2026-08-14
- Zhipu's GLM 5.3 drops overnight with dramatic coding gains — bindureddy · 2026-08-14
- Alibaba Qwen teases: less than 2 hours to say hi, it's almost time — Alibaba_Qwen · 2026-08-14
- Gemini 3.7 Flash drops 'minimal' thinking level, matching 3.1 Pro — PleasantSir9581 · 2026-08-14
- Xiaohongshu Open-Sources dots3-note: Perfect 42/42 at IMO, Apache 2.0 — eyishazyer · 2026-08-14