
What Models Can RTX 4090 Run? An Honest 2026 Guide to 24GB VRAM
An RTX 4090 runs 8B and 14B models with room to spare, 32B at Q4 up to about 16K context, and no 70B on a single…
AI, hardware, and the tools I actually use. The notes are all here.

An RTX 4090 runs 8B and 14B models with room to spare, 32B at Q4 up to about 16K context, and no 70B on a single…

AI Max 395 vs DGX Spark compared with real llama.cpp numbers: similar writing speed, Spark reads prompts 2 to 5 times faster, and the AMD boxes…

A living RAM and SSD price tracker: dated DDR4 and DDR5 spot, contract and retail prices, a 12-month per-GB chart, what is driving prices, forecasts from…

A Claude Code 429 can be your Pro/Max session limit, your API key's rate limit or a monthly spend cap, and each needs a different fix.…

A skill teaches Claude how to do something; MCP connects it to an outside service. What Anthropic's docs say about Claude Code skills vs MCP, hooks…

Repeated 529 Overloaded errors in Claude Code mean Anthropic is out of capacity for that model. It doesn't count against your quota. How it differs from…

Ultrathink was reported dead in January 2026, came back in 2.1.68, and since Claude Code 2.1.311 it runs a terminal turn at xhigh. What each effort…

OpenWrt 主题 Lumen 是我做的 ucode LuCI 界面:深浅色、侧栏、实时仪表盘和手机布局。v1.0.1 支持 25.12.6 APK 和 24.10.8 IPK,附签名验证和回退步骤。

The Anthropic IPO S-1 is out. What it confirms, where the $80.7 billion loss really comes from, why Class A shares carry almost no vote, how…

"Selected model is at capacity. Please try a different model." isn't your usage limit. What the Codex error means in the source code, what OpenAI staff…

Rubin vs MI455X specs from NVIDIA's and AMD's own pages. NVIDIA's Rubin page quietly went from 22 to 19.2 TB/s per GPU in early September. Full…

GPU rental prices keep rising, but IREN, CoreWeave and Nebius fell 30% to 47% in a month. The 1-month drop vs the drop from the high,…