{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreifh3tjzt36rk2zsjxbu6yvgdwettjwrfg56iwidxilllroau2um4q",
"uri": "at://did:plc:ajcrkmnlj6rxdk7rltijv227/app.bsky.feed.post/3mmjh7nj36tz2"
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreihwdoy5g26ba26zjzx7aq76xi3dl5yleaw3zrotci7bo2godopruu"
},
"mimeType": "image/png",
"size": 1945970
},
"path": "/tech-industry/artificial-intelligence/enthusiast-runs-1-trillion-parameter-llm-from-768gb-of-intel-optane-dimm-memory-sticks-local-kimi-k2-5-install-achieved-roughly-4-tokens-per-second",
"publishedAt": "2026-05-23T11:20:00.000Z",
"site": "https://www.tomshardware.com",
"tags": [
"Artificial Intelligence",
"Tech Industry"
],
"textContent": "A Redditor has caused a stir by coaxing a workstation build using Optane PMem DIMMs as RAM to run a 1-trillion parameter LLM.",
"title": "768GB of cheap Intel Optane DIMM memory sticks used to run 1-trillion-parameter LLM on a system with a single GPU — local Kimi K2.5 install achieved roughly 4 tokens per second"
}