![AB Notes / AI × 1C × [ ... ]](https://cdn4.telesco.pe/file/J-Qorf1ymnWmWsAXivJh7L4_xVbL5dU6nXoc2ZeqjNf41fVCh2DruuszigdoXNSEry9CAYxQvs2RTzfX5l6GePPD2JY7ckWdP7oFuSr57vV7ieYPSlnePU_sAnE2lL6ULbMgBd-RZscukWbjiPXemwGW99quO6zBOvU1gAbTUnlvtXJLc8t694ELUCcFjMMX281jd_tsO_i9268lWPRnSvHIS6NHljIcUp6KILmt86CxO6iBq3p1Q0l_eMnmJMSE6CsZABBW2ONCyPbckSCG2WGJCBSyKYwYpNN-fvRmPfmBVp7cQqFe6sKiA9NvfA2-bFrY5Ju7mZm4siw9dJm6fQ.jpg)
DeepSeek v4🔥
Бенчмарки, карточка модели: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro#deepseek-v4-pro-max-vs-frontier-models
По бенчам SOTA среди открытых моделей, чуть хуже чем Opus-4.6, GPT-5.4
API:
https://api-docs.deepseek.com/quick_start/pricing
Контекст: 1 млн
V4-Flash: in $0.14 / cache hit $0.028 / out $0.28
V4-Pro: in $1.74 / cache hit $0.145 / out $3.48
Post #69
363
- 🔥 8