[REVISION HISTORY]

Apple M5 and M6 chip development and performance

Updated 15 times since CLSTR started tracking revisions of this situation.

What changed

2026-09-24 22:38 UTC → 2026-09-25 00:15 UTC · added ~~removed~~

Apple has launched its 2026 desktop lineup, featuring Mac mini and Mac Studio computers powered by M6 and M5 series chips. The new Mac mini utilizes the first 2nm M6 chip, which includes a dual 16-core Neural Engine designed to provide up to “four times the AI performance of its predecessor” and 40 percent faster CPU performance. This hardware targets the enterprise AI market by enabling high-performance local tasks, such as running large language models, to reduce reliance on cloud-based services. While the chip offers significant gains, the base price has increased to $899 due to rising costs for RAM and storage. Technical analysis has raised concerns regarding the 256GB M6 Mac mini; its single-chip NAND configuration may fail to achieve the “2x storage speed increase” seen in higher-capacity models, resulting in slower read and write speeds. ~~Furthermore,~~ Recent reports confirm that the SSD module is “no longer listed as a ~~user-replaceable part,” moving toward soldered~~ replaceable part” for the M6 and M5 Pro versions. While the M4 generation featured removable NAND storage modules that ~~prevents the~~ allowed for third-party ~~upgrades possible with~~ upgrades, the ~~M4 generation.~~ latest models have the storage soldered directly to the logic board. Specialists such as Expand Mac mini indicate that the previous modular design was a rare exception. Consequently, users are advised to select their required storage capacity at the time of purchase, as internal memory cannot be expanded after assembly. The high-end Mac Studio models are equipped with M5 Max and M5 Ultra chips. The M5 Ultra variant can support up to 512GB of unified memory with 1.2TB/s bandwidth. Technical reviews of the M5 Ultra highlight significant performance leaps over the M3 Ultra, including approximately 30% faster single-core and 50% faster multi-core CPU speeds. ~~Apple has highlighted the ability to cluster multiple Mac Studios using Thunderbolt 5 and RDMA for distributed AI inference, potentially delivering three times the performance of a single machine.~~

Versions

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.