The new SingleStoreDB tool truck features numerous performance and scalability improvements, including Flexible parallelism, which allows queries to use all cores on a machine regardless of the number of partitions. The release also includes Spilling for hash group-by, select-distinct and count-distinct operations, which writes part of the in-memory hash table to disk if it exceeds available memory, allowing queries to run even if they hit the memory limit. Additionally, SingleStoreDB upgraded its LLVM code generation framework to version 10, resulting in improved query performance. The tool truck also introduces Materialized CTEs, row-level decoding for string columns, faster upgrades via reduced effort for codegen, and a new SET statement for user-defined session variables. Other features include Matching expressions to computed columns, VECTOR_SORT and other vector functions, ISNUMERIC, SELECT FOR UPDATE – now for Universal Storage, SECRET(), and various storage improvements such as backup with split partitions for unlimited storage DBs and drop milestones.