Cerebras announces CS-4, its fourth-generation wafer-scale system
Cerebras unveiled CS-4, built on three Wafer Scale Engine 3 Turbo processors on a new modular Nexus rack, claiming up to 30x faster inference than GPUs and up to 2x faster than CS-3; pricing is undisclosed and sold via enterprise contract.
CS-3 (WSE-3, announced March 2024) was Cerebras's newest enterprise compute system, sold via direct enterprise contract with no public pricing.
CS-4 (announced Aug 18, 2026) becomes Cerebras's newest system — three WSE-3 Turbo processors on a redesigned modular 'Nexus Platform' rack, claimed up to 30x faster than GPU systems and up to 2x faster than CS-3, with up to 10x more throughput per watt. No public pricing disclosed; first shipments described as beginning Q3 2026.
Proof of change
Cerebras's pricing pages, as we captured them on two dates.
These values appear in only one of the two captures. That can mean a page-layout difference rather than a price move — read the images, not just the list.
Showing the whole page as captured — scroll either panel, or open it at full size.
Show 2 other pages we compared
Showing the whole page as captured — scroll either panel, or open it at full size.
Showing the whole page as captured — scroll either panel, or open it at full size.
- These prices come from our capture alone — they were not confirmed against an independent second source.
- 2 pages had no counterpart in the earlier capture and are not shown.
Both images are our own captures, taken on the dates shown. The pixels are unmodified — nothing is retouched, and where a region is highlighted the marker is drawn over the image, not into it. Open either capture to see it at full resolution.
Cerebras announced CS-4 on August 18, 2026, its fourth-generation compute system and the newest addition to its enterprise hardware line alongside CS-3. The system pairs three new Wafer Scale Engine 3 Turbo processors with a completely redesigned rack architecture Cerebras calls the “Nexus Platform” — a modular design the company says cuts component count by roughly half and reduces wafer-to-wafer interconnect latency to as low as 2 microseconds, enabling more than 1,000 tokens/second on models exceeding 10 trillion parameters.
Cerebras claims CS-4 delivers up to 30x faster inference than production GPU systems and up to 10x more throughput per watt than CS-3, with up to 2x faster raw performance than its predecessor. As with CS-3, no public pricing is disclosed — CS-4 is sold exclusively through direct enterprise sales engagement, with Cerebras stating first shipments begin “this quarter” (Q3 2026). The announcement does not affect pricing or packaging on Cerebras’s separate, self-serve Inference API product line (Free Trial, Developer, Enterprise tiers).
Cerebras announced CS-4, built from three new Wafer Scale Engine 3 Turbo processors on a redesigned modular rack ("Nexus Platform"). Cerebras claims up to 30x faster inference than GPU systems, up to 10x more throughput per watt and up to 2x faster performance than CS-3, and wafer-to-wafer interconnect latency as low as 2 microseconds. No pricing was disclosed — CS-4 remains an enterprise-contract, "contact us" product like CS-3 — with first shipments described as beginning "this quarter" (Q3 2026).