DeepSeek Doubles Down on Low-Cost AI With V4-Flash Update
The Chinese AI startup is strengthening its position in the global AI race as developers compete on performance, efficiency, and affordability.
News
- Mubadala Eyes Up to $6.3 Bn Investment in Japan AI Data Center
- Alibaba Plans Revenue-Sharing Model for Next Open-Weight AI Release
- AI Dispatch: Global Reorder and a Region on the Rise
- Ajman Targets 100 AI Initiatives Under Government-Wide Program
- JPMorgan CEO Leads 40-Company Effort to Tackle AI Risks
- Identity, More Than Software Flaws, Is Driving Today's Costliest Cyberattacks
Chinese AI startup DeepSeek has released an updated version of its V4-Flash model, positioning it as one of the most cost-efficient AI systems among major global models, according to research firm Artificial Analysis.
The DeepSeek V4-Flash 0731 represents an improvement over the previous version, with Artificial Analysis describing it as a significant step forward. The model achieved a score of 50 out of 100 on the firm’s Intelligence Index, matching Google’s Gemini 3.6 Flash in the ranking.
Cost remains DeepSeek’s main differentiator. Artificial Analysis estimated the model’s average cost per benchmark test at around $0.03, compared with $0.86 for Moonshot AI’s Kimi K3, $1.86 for OpenAI’s GPT-5.6 Sol, and $3.15 for Anthropic’s Claude Fable 5.
DeepSeek V4-Flash retains a 1 million-token context window, with 284 billion total parameters and 13 billion active parameters during inference. The model is available through DeepSeek’s API and is part of a broader push by Chinese AI companies—including Alibaba and Moonshot AI—to compete through high-performance, cost-efficient models.
Reportedly, the AI lab is gearing up for a potential IPO, and the model launch is a strategy to regain momentum by delivering a model suited to its best capabilities — an ultra-low-cost AI alternative.
According to Artificial Analysis, the new version is “a significant step up from the previous generation, DeepSeek V4 Flash (40).”
Despite OpenAI cutting GPT-5.6 token prices by up to 80%, DeepSeek’s new model still comes out nearly 60% cheaper per task for similar tasks. “DeepSeek V4 Flash 0731 used ~206M output tokens to run the Intelligence Index, against ~234M for the previous DeepSeek V4 Flash,” the report read.
The startup is also speculated to launch a more advanced model, called the V4-Pro.
Recently, Chinese AI players have been dominating the landscape with advanced models. These include Alibaba with Qwen3.8 and Moonshot’s Kimi K3.
Available through DeepSeek’s first-party API, the model retains a 1M-token context window, with 284B total parameters and 13B active at inference time.
MIT Sloan Management Review Middle East invites you to the second edition of AI Research Forum — “Agentic AI: From Mandate to Momentum” taking place on 22 October 2026 in Dubai.To partner, speak, or attend AIRF, click here.
