DeepSeek V4 Revolutionizes AI Coding with Affordable Pricing
In brief
- DeepSeek has launched its V4 model, an open-source AI system available under the MIT license.
- Priced at $0.30 per million output tokens, it’s significantly cheaper than competitors like Claude Opus 4.7 ($25) and GPT-5.5 ($30).
- The model scored 80.6% on SWE-bench Verified, just 0.2 points behind Claude Opus 4.6.
- Its efficient architecture, including a 1.6-trillion-parameter MoE and optimized inference processes, makes this pricing sustainable without being a loss leader.
- While self-hosting requires substantial resources, the model’s capabilities in coding tasks like LiveCodeBench and Codeforces are unmatched, challenging the dominance of closed models.
- However, concerns about benchmark transparency and data governance remain.
- This release resets the price floor for high-quality coding AI, potentially forcing competitors to adjust their strategies or enhance unique features.
Terms in this brief
- MoE
- Mixture of Experts — a technique where a large neural network is divided into smaller, specialized sub-networks (experts) that work together to make decisions. This approach improves efficiency and scalability in AI models.
- SWE-bench
- A benchmark designed to evaluate an AI's ability to fix real bugs found in actual open-source software projects on GitHub. It assesses the practical coding skills of AI systems by presenting genuine engineering challenges, rather than theoretical exercises.
Read full story at Hacker News →
More briefs
Tesla Robotaxis Crash Twice Since July 2025
Tesla’s robotaxis have crashed at least twice since July 2025 while being remotely driven by operators. Both incidents occurred in Austin, Texas, at low speeds with safety monitors onboard and no passengers. The crashes include one where the vehicle hit a metal fence after moving forward from a stopped position and another where it scraped a fender after contacting a construction barricade at 9mph. Unlike other companies, Tesla previously redacted crash details, but now provides full narratives for all 17 recorded incidents since last year. While most crashes involve being hit by others, two instances saw the robotaxis damaging their own mirrors. These events highlight challenges in autonomous driving systems and raise questions about safety as Tesla expands its robotaxi network.
Japan's Robotic Wolves Help Fight Bear Attacks
A Japanese company is struggling to meet high demand for its robotic wolves designed to scare off wild animals. Ohta Seiki, based in Hokkaido, has received over 50 orders this year-more than usual-and is asking customers to wait two to three months. The surge follows a record number of bear attacks and sightings, with nearly 14,600 bears culled last year alone. The "Monster Wolf" device uses motion activation, LED lights, and recorded sounds to deter animals from entering agricultural land or urban areas. The company is also developing an AI camera-equipped version on wheels for chasing animals, as well as a smaller handheld model for hikers. This innovative approach aims to protect people and wildlife while addressing the growing challenge of human-bear coexistence.
AI Shifts Job Market Favoring Older Workers
A new survey reveals that companies adopting artificial intelligence are cutting junior roles and favoring mid- to senior-level employees. Over 40% of CEOs plan to reduce entry-level positions while only 17% intend to expand them, a significant shift from just a year ago. AI's ability to handle tasks like coding and sales analysis makes experienced workers more valuable due to their judgment skills and problem-solving abilities. However, this trend risks creating future shortages of young talent, as companies rely increasingly on AI agents. Despite this shift, older workers remain vulnerable as no role is entirely secure in the evolving job market driven by technology.
AI-Generated Short Dramas Are Taking Over Global Streaming
A new wave of short dramas, entirely created using AI without any human actors or CGI, is sweeping through platforms like DramaWave and ReelShort. These ultra-short, melodramatic shows often feature intense emotional scenes and are designed for quick smartphone viewing-perfect for endless scrolling. The industry, which began in China in 2018, has grown tremendously, with $6.9 billion in revenue in 2024, surpassing traditional box office earnings. Now, Chinese companies are expanding globally, translating shows for international audiences and dominating markets like the U.S., where these apps generate about half of their revenue. AI is revolutionizing production, enabling 470 new shows to be released daily in January alone. Studios are cutting costs by shrinking film crews and relying on AI for everything from scriptwriting to editing, reducing production timelines from three months to less than a month. This shift is transforming entertainment, making it faster, cheaper, and more accessible worldwide.
AI Data Centers Push Demand for Ideal Power's B-TRAN Technology
Ideal Power Inc.'s CEO, David Somo, revealed that the company is seeing growing opportunities in AI data centers and energy infrastructure. The firm’s sales pipeline has expanded from around $200 million to over $300 million this year, split evenly between automotive and data center applications. A key driver is the increasing power demands of AI systems. Historically, data center racks used 10-50 kilowatts, but new AI-focused systems now require 100-150 kilowatts, with future expectations surpassing 500 kilowatts per rack. Ideal Power’s B-TRAN semiconductor technology is well-suited for these transitions, as data centers shift from traditional AC systems to modern DC architectures. The company is also expanding in Asia, targeting markets like energy storage and smart industrial buildings. Looking ahead, Somo emphasized the potential for solid-state transformers and collaborations with hyperscalers to further demonstrate B-TRAN’s value. Ideal Power aims to capitalize on these trends by advancing customer testing and securing initial revenue opportunities later this year.