DeepSeek V4-Flash Tops OpenRouter, Costs 1% of Rivals

DeepSeek V4-Flash (version 0731) is officially live, rapidly claiming the top spot for call volume on the OpenRouter platform thanks to its extreme cost-performance ratio. Its single-day Token consumption exceeded 8 trillion, surpassing OpenRouter's total daily platform average. The model set a new SOTA record for cost-to-accuracy ratio in benchmarks, signaling that LLM competition is shifting from "single-turn intelligence contests" to "cost-efficiency calculations for long tasks."

Confirmed

Performance & Cost: In the WeirdML evaluation, DeepSeek v4 Flash scored 57.1% and 63.0%, beating the recently price-reduced GPT 5.6 Luna. Data from independent evaluation firm Artificial Analysis shows that in complex, real-world workloads, its cost is only $0.03, compared to a staggering $3.15 for Claude Fable 5.
Market Reception: The total call volume for domestic models has continuously surpassed the combined total of US closed-source models for multiple weeks. A massive number of overseas developers have proactively migrated their production environments to this model.
Tangible Pricing: According to Quantum Bit (量子位), the model's price is incredibly low. For example, running a 3D shooter game costs about 7 cents per session, and a game of CS about 50 cents. Furthermore, overseas proxies are stacking subsidies, leading to quotes in the single-cent range.

Why It Matters

With its ultra-low price and outstanding Agent potential, V4 Flash provides developers with a far more economical choice, fundamentally reshaping the global LLM market's pricing system and developer usage patterns.

2026-08-03 ~ 2026-08-05 · 11 related posts

Primary sources

DeepSeek V4 Costs 1% of Claude: China's AI Price War Disrupts the Market — SirBoboGargle · 2026-08-04 16:19 (UTC+08:00)
DeepSeek v4 Flash Sets New Cost/Accuracy SOTA on WeirdML — teortaxesTex · 2026-08-03 20:20 (UTC+08:00)
DeepSeek V4 Flash Burns 8T Tokens Daily, Reshaping Agent-Era Model Pricing — APPSO · 2026-08-03 16:26 (UTC+08:00)

[source] DeepSeek V4 Flash Burns 8T Tokens Daily, Reshaping Agent-Era Model Pricing — APPSO · 2026-08-03
DeepSeek-V4-Flash tops OpenRouter, reshaping LLM pricing with extreme cost-efficiency — 创业邦 · 2026-08-03
[source] DeepSeek v4 Flash Sets New Cost/Accuracy SOTA on WeirdML — teortaxesTex · 2026-08-03
DeepSeek V4 Flash tops the Vals Index above 60 at 35x lower cost — zephyr_z9 · 2026-08-04
DeepSeek V4-Flash is being used for 3D games at cents-level costs — 量子位 · 2026-08-04
[source] DeepSeek V4 Costs 1% of Claude: China's AI Price War Disrupts the Market — SirBoboGargle · 2026-08-04
Stop Burning Money on Opus: The DeepSeek V4 Flash Fix — PrajwalTomar_ · 2026-08-04
DeepSeek Dubbed the 'Model T' of the AI Era for Its Aggressive Pricing — FuSheng_0306 · 2026-08-04
DeepSeek V4 Flash Costs 1/100th of Top Models, Set to Trigger Autonomous Agent Wave — Imaginary_Dinner2710 · 2026-08-05
DeepSeek Flash's performance is mind-boggling,网友直呼“不合理” — yacineMTB · 2026-08-05
Developer Marvels at DeepSeek's Pricing: Indefinite Use for Almost Zero Cost — yacineMTB · 2026-08-05