{"id":6012,"date":"2026-09-22T01:41:24","date_gmt":"2026-09-22T01:41:24","guid":{"rendered":"https:\/\/www.torck.io\/?page_id=6012"},"modified":"2026-09-22T01:41:24","modified_gmt":"2026-09-22T01:41:24","slug":"rtx-4090","status":"publish","type":"page","link":"https:\/\/www.torck.io\/en\/llm-finder\/rtx-4090\/","title":{"rendered":"Plan LLM deployment on RTX 4090"},"content":{"rendered":"
\n

Which LLMs fit on an RTX 4090<\/h2>

As of the data snapshot of

The first test candidate is Devstral-Small-2-24B-Instruct-2512 in Q4_K_M with llama.cpp. This configuration needs up to 18.1 GiB, estimated at 37 to 52 tokens per second. Runtime status llama.cpp: Build b10964 from Sep 14, 2026 (release notes<\/a>). Suitability: no independent data. Memory and speed are calculated, not measured. Only a test with your own questions shows whether a model suits your task.<\/p>

Go to the full result<\/a> \u00b7 Calculation method<\/a><\/p><\/section>

torck LLM Finder<\/span>

Your requirements<\/h2>400 computable models \u00b7 1,092 configurations \u00b7 93 hardware options \u00b7 Data snapshot

Straight to the result<\/a><\/p>

\n
Start with<\/legend>

Select up to four configurations<\/h3>

Nothing selected yet. Search above for a model to compare configurations.<\/p><\/div><\/div>\n

\n

Hardware undecided: start with a 48 GiB planning profile<\/a><\/p>\n

Memory and serving assumptions<\/summary>