tinybox owner: $11k-$22k local AI boxes beat buying GPUs outright
CedricMakes · x · 2026-11-02
CedricMakes shares hands-on experience with tinygrad's tinybox: an OG Green 6x 4090 box ran small non-LLM models 24/7 for weeks at home and in a colo. He plans to buy a newer box for local SOTA open models if closed models get more restrictive, noting GPU market prices are so crazy a whole tinybox is sometimes cheaper than just the GPU. tinygrad quotes $22k for a tinybox red flash (GLM-5.3-Flash) and $11k for a 4x9070XT tinybox red (Qwen3.8 Flash Next), urging people to stop fiddling with single-card setups.
More from Infra
- Demis Hassabis: AI data centers should be beautiful — Recent_Fox4339 · 2026-11-02
- CVSS 9.8 Linux kernel bug CVE-2026-98432 allows remote takeover; Ubuntu still unpatched — NathanpmYoung · 2026-11-02
- RTX 4090 volatile unlock boosts FLUX.1-dev inference 64% to 214 TFLOPS — BrightRaiderDev · 2026-11-02
- Ava: an open-source real-time voice avatar that runs fully on-device in your browser — victormustar · 2026-11-02
- Anthropic and OpenAI would weather a compute glut best, investor argues — bookwormengr · 2026-11-02
- Ava: in-browser local AI avatar runs LFM3-Omni via WebGPU, no cloud needed — paf1138 · 2026-11-02