Background

In early August, the LLM competition axis shifted from "who scores higher" to "cost per Intelligence Index point." OpenAI cut GPT-5.6 Terra and Luna prices by 20% and 80% respectively on July 30, dragging the industry's focus to this new metric. DeepSeek's V4 Flash 0731, released as a public-API test on July 31, lands directly on that axis (Huatai Securities research note, Aug 5).

Core Numbers: 50 on the Index, 60% Cheaper Per Task

The Artificial Analysis evaluation published July 31 lays out V4 Flash 0731's headline numbers:

  • Intelligence Index v4.1: 50 points — 10 points above the previous V4 Flash (40), 6 points above V4 Pro (44), and only 1 point behind GPT-5.6 Luna (max, 51) (Artificial Analysis article).
  • Cost-per-task is ~60% lower than GPT-5.6 Luna — even after Luna's 80% price cut, this 60% gap holds (same Artificial Analysis source).
  • Agentic eval gains are real: GDPval-AA v2 jumped from 1189 to 1559, Terminal-Bench 2.1 up 17 points to 79%, τ³-Bench Banking up 8 points to 31%.
  • Hallucination rate down 12 points to 84% — sitting in the same tier as GPT-5.6 Terra (85%) and Mistral Medium 3.5 (82%).
  • Total output token usage down 12% — the model became more concise while scoring higher.

Architecture & Pricing: No New Skeleton, Just Post-Training

V4 Flash 0731 didn't change any architecture. It inherits V4 Flash preview's pure-text MoE design:

  • 284B total parameters / 13B active at inference;
  • 1M token context window, max output 384K tokens;
  • 43-layer Transformer hybrid attention (first two layers are pure sliding window, later layers alternate CSA / HCA compressed attention);
  • mHC super-connection + Muon optimizer (same-series design);
  • Official API pricing /bin/bash.14 / /bin/bash.28 per 1M input/output tokens — unchanged from V4 Flash;
  • 98% cache hit discount — far above the industry-standard 90% (direct quote from the Artificial Analysis article).

On the post-training side, V4 Flash 0731 mostly reworked agent capability: DeepSWE jumped from 7.3 (preview) to 54.4 (a 7.5× leap), Terminal-Bench hit 82.7, approaching Claude Opus 4.8's 85 (aitoollab review).

Where V4 Flash 0731 Sits in the China-AI Landscape

Combining the Huatai Securities research note and Artificial Analysis's evaluation:

  • Capability ceiling: Kimi K3's 57 points is still the open-weights frontier for Chinese models.
  • V4 Flash 0731 sits in the 50-point tier — 1 point behind GLM-5.2 (51), tied with Gemini 3.6 Flash (50).
  • Value floor: V4 Flash 0731's blended price is ~/bin/bash.06/1M tokens, average cost-per-task ~/bin/bash.03 — about 65% and 57% lower than Luna respectively (same Huatai source).
  • Full weights "expected in the coming weeks" (direct quote from Artificial Analysis); Unsloth's GGUF quant is already runnable locally.

Personal Take

V4 Flash 0731's real signal isn't the 50-point headline — it ties Gemini 3.6 Flash and only trails Luna by 1 point, which is a narrow margin. The real signal is that without changing architecture and only doing post-training, it pulled agent capability up to where it can stand toe-to-toe with flagship models. This means DeepSeek has pushed the marginal return of "weights → capability" to a very high level, and the iteration speed will only accelerate from here.

The more under-rated number is the cache hit discount: 98% vs the industry's common 90%. That 8-point gap is order-of-magnitude impactful in long-context agent scenarios — in a 1M-token conversation where 90% hits cache, DeepSeek only charges for 100K tokens, not 1M. OpenAI's 80% price cut is "make the user see cheapness"; DeepSeek's design is "make cache actually cheap." The two philosophies are completely different.

China's open-weights models have now formed a dual-track pattern — Kimi K3 owns the capability ceiling, V4 Flash 0731 owns the value floor — which is a perfect mirror of OpenAI's "Sol preserves margin + Luna wins market share." What to watch next: when V4 Flash 0731's full weights open up, can the local-deployment ecosystem amplify that 60% cost gap even further?