DeepSeek V4 Pro benchmarked against Claude: only 5% worse at 4,500% lower cost

DeepSeek's updated V4 Pro model scored only 5% below Anthropic's flagship Claude in benchmarks, despite costing roughly 4,500% less. An earlier April preview had landed 18 points behind Claude, but DeepSeek's own numbers for the finished model tell a far more competitive story. The result poses a serious challenge to pricier Western AI rivals on value grounds.