Latest 6G and AI news, analysis, and industry updates.
DeepSeek's August 16 move to UTC-based peak/off-peak pricing for V4-Flash and V4-Pro pushes cache-hit costs up by more than 10x during busy hours, ending the era of near-free caching and forcing agent builders to actually schedule their workloads.
Kimi K3's 2.8T parameters and 1M context made headlines, but the deeper story is bargaining power: planned open weights turn frontier capability into a deployable second source. OpenCode's same-day integration, paired with its warning that inference costs remain undiscounted, shows the real bottleneck has shifted from access to affordability.