DeepSeek V4 Flash Scores 50 on Intelligence Index, Undercutting GPT-5.6 by 60%
solyarisoftware · x · 2026-08-01
According to Artificial Analysis, DeepSeek V4 Flash (0731) achieved an intelligence index score of 50, representing a massive leap over its predecessor and tying with Gemini 3.6 Flash.
While maintaining extremely low API pricing, the model even outperforms the previous DeepSeek V4 Pro. Driven by a 98% cache hit rate, DeepSeek's cost per task remains about 60% lower than that of GPT-5.6 Luna, even after OpenAI's 80% price cut today.
More from Models
- Google's Astra Model Proves Existence of Nonsofic Groups, Solves Multiple Math Problems — xeophon · 2026-08-01
- OpenAI Slashes GPT-5.6 Luna API Price by 80%, Rivaling Opus 5 at 1/17 Cost — gabrielchua · 2026-08-01
- Claude Opus 5 Called Unusable, Sonnet 5 Praised as Most Reliable Model — seanmcdonaldxyz · 2026-08-01
- Did Frontier Models Adopt the Unlimited OCR Tech? Community Weighs In — Wise_Stick9613 · 2026-08-01
- Users Report Claude Opus Has 'Weird Psychiatric Tone' and Poor German Grammar — MarcJSchmidt · 2026-08-01
- DeepSeek Builds Working Game for Just $0.07, Making Coding Too Cheap to Meter — rand_longevity · 2026-08-01