In a dramatic escalation of the AI arms race, OpenAI has released GPT-5.5 while Chinese lab DeepSeek has published details on its V4 model, intensifying the global competition for computing power.
OpenAI's latest iteration, GPT-5.5, comes with a doubled price tag compared to its predecessor, reflecting the soaring costs of training and inference. Early benchmarks show significant improvements in reasoning and coding tasks, though some critics argue the gains are incremental given the price hike.
Meanwhile, DeepSeek's V4 paper reveals a model that rivals GPT-5.5 on several metrics while running on domestically produced Chinese chips—a strategic move amid export restrictions on advanced semiconductors. The development signals China's growing self-sufficiency in AI hardware and software.
The news has reignited debate over the "compute war," with both US and Chinese firms racing to secure data center capacity and next-generation chips. Analysts warn that the demand for compute is outstripping supply, potentially creating a bottleneck for future AI progress.
In related developments, the video also covers Mythos's marketing push, experiments with vibe-coded games using GPT Image 2, and 50 data points that go beyond headlines. The scarce compute era is expected to define the next phase of AI development.