DeepSeek V4 Flash 0731 Scores 50 on Intelligence Index, Closing Gap with GPT-5.6
ArtificialAnlys · x · 2026-07-31
Artificial Analysis released its latest evaluation of DeepSeek V4 Flash 0731. The model scores 50 on the Intelligence Index, a 10-point jump that puts it 6 points ahead of DeepSeek V4 Pro, landing perfectly on the Pareto frontier for intelligence vs. cost.
Key Data & Highlights:
- Extreme Cost-Effectiveness: Thanks to a 98% cache hit discount, its cost per task is about 60% lower than the similarly intelligent GPT-5.6 Luna.
- Massive Agentic Gains: Elo rating on GDPval-AA v2 surged from 1189 to 1559, with significant improvements in Terminal-Bench and other tests.
- Token Efficiency: Output tokens used for the Intelligence Index dropped 12% compared to its predecessor.
- Reduced Hallucinations: The AA-Omniscience Hallucination Rate fell by 12 points to 84%, driving overall improvement while maintaining a 284B parameter size.
The model retains a 1M token context window, with full weights expected to be released in the coming weeks.
Related event: DeepSeek V4-Flash: Agent Leap, Open Source, Ultra-Low Price(100 posts)→
More from Models
- 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
- NVIDIA ships FP4-quantized Qwen3.8-Flash for local use, trending on Hugging Face — nvidia · 2026-10-16
- RoboVista benchmark: Gemini-4 Argon hits 81.9% on robotics understanding — berkeley_ai · 2026-10-16
- Dev reports decision model jev beats gpt-6-luna on quality in real training pipeline — genmon · 2026-10-16