Summary
The Register reported that CS-4 uses the WSE-3T/Turbo, which Cerebras says doubles compute, memory-fabric and I/O bandwidth while retaining its 5nm process and 4 trillion transistors. Cerebras claims up to 4,400 tokens per second per user on gpt-oss-120b, with first systems due later this quarter.
Why it matters
The system shows how inference hardware is moving towards modular rack-scale architectures that compete with GPU-based infrastructure on throughput and deployment density.
Related coverage
- Relativity Networks raises $22 million for faster data-centre fibre
- DataVita secures £300 million for Scotland AI Growth Zone
- Google adds interactive study tools across Search and Gemini