{
  "id": 73712,
  "title": "DeepSeek’s new AI model undercuts rivals on running costs",
  "url": "https://urgent.news/2026/08/03/deepseeks-new-ai-model-undercuts-rivals-on-running-costs",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-03T06:31:30.000Z",
  "source": {
    "name": "Free Malaysia Today",
    "slug": "free-malaysia-today",
    "url": "https://www.freemalaysiatoday.com/category/business/2026/08/03/deepseek-s-new-ai-model-far-cheapest-of-well-known-models-to-run"
  },
  "original_language": "en",
  "account": "Beijing, China - A newly released AI model by Chinese startup DeepSeek, dubbed V4-Flash, is proving to be a formidable competitor with its ultra-low running costs, according to a recent report by Artificial Analysis, a leading research firm. The model has demonstrated performance comparable to Google's Gemini 3.6 Flash, yet it significantly outperforms rivals such as OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5 in terms of cost efficiency. Specifically, DeepSeek's V4-Flash charges mere pennies per million input tokens and even less per million output tokens, making it more than 100 times cheaper to run than Anthropic's Claude Fable 5.\n\nThis cost advantage is particularly noteworthy given that DeepSeek's earlier model, R1, had ignited a global sensation in early 2025, sparking a sell-off in global technology stocks and sparking debates on the exorbitant AI expenditures of U.S. companies. The latest model, V4-Flash, is priced at a mere $0.14 per million input tokens and $0.28 per million output tokens, with Artificial Analysis estimating an average cost of $0.03 per test. This is a stark contrast to the prices of competitors, which range from $0.86 for Moonshot AI's Kimi K3 to $1.86 for OpenAI's GPT-5.6 Sol and $3.15 for Claude Fable 5.\n\nThe pricing of an AI model is not merely determined by its headline cost, but also by the volume of data it processes and generates to accomplish a task. A model with a lower price tag might still prove to be expensive if it necessitates a significantly larger number of steps to produce an answer. DeepSeek's V4-Flash model has scored 50 out of 100 on Artificial Analysis's Intelligence Index, which is a composite score derived from nine benchmarks encompassing coding, reasoning, and workplace-style assignments. This score places the model on par with Google's Gemini 3.6 Flash and one point behind Meta's Muse Spark 1.1 and Z.AI's GLM-5.2.\n\nDespite these strong performances, Moonshot AI's Kimi K3 has edged out DeepSeek with a score of 57, while Anthropic's Claude Opus 5, Fable 5, and OpenAI GPT-5.6 have scored even higher, scoring nine or more points above DeepSeek's V4-Flash. Furthermore, DeepSeek is reportedly gearing up for the release of an even more powerful version of its model, the V4-Pro, though the timeline for this release remains undisclosed. On another front, Alibaba recently unveiled its largest and most capable AI model, the Qwen3.8-Max, which, when compared to its domestic rival Moonshot AI's recent launch, demonstrates a comparable level of sophistication.",
  "summary": "DeepSeek officially released its V4-Flash model on Friday, its latest attempt to regain momentum by doing what it is best known for - offering ultra-low-cost AI alternatives.",
  "key_points": [
    "DeepSeek's V4-Flash AI model offers ultra-low running costs.",
    "V4-Flash costs $0.14 per million input tokens, $0.28 per million output tokens.",
    "V4-Flash scores 50/100 on Artificial Analysis's Intelligence Index."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}