Can I Run Codestral on NVIDIA GeForce RTX 5090?

Name: LLM Configurator — GPU VRAM Checker
Author: LLM Configurator

Superseded model. Codestral has been superseded by Devstral. This page is kept for reference; the newer family is a better starting point. View Devstral →

作者： Jakub Rusinowski · 最后更新： 2026年7月12日

Yes, comfortably — you'll have ~18.6 GB of headroom running Codestral 22B at Q4_K_M (13.4 GB, ~86 tok/s (est.)) with room for up to 32K context.

See what else this hardware can run →

购买此硬件 NVIDIA GeForce RTX 5090 32GB — 32 GB VRAM · 575 W board power 立即云端部署 RunPod 上的 RTX 5090

或在 Vast.ai 比较

作为亚马逊联盟成员，我们从符合条件的购买中获得收入。云 GPU 链接为推荐链接——我们可能获得佣金，您无需额外付费。

联盟营销声明: 本页部分链接为联盟推广链接——如果你通过它们购买，LLM Configurator 可能会获得佣金，而你无需支付任何额外费用。作为亚马逊联盟成员（Amazon Associate），LLM Configurator 会从符合条件的购买中获得收益。

NVIDIA GeForce RTX 5090 32GB

32 GB VRAM · 575 W board power

2026年价格波动较大——请以当前商品页价格为准。

在亚马逊查看价格

NVIDIA GeForce RTX 5090 Specs

VRAM	32 GB
Memory Bandwidth	1792 GB/s

Codestral 22B on the NVIDIA GeForce RTX 5090: VRAM by quantization

Quant	VRAM needed	Fits 32 GB?	Max context	Whole PC
F16	46.1 GB	✗ No	—	—
Q8_0	25.3 GB	✓ Yes	32K	—
Q6_K	19.9 GB	✓ Yes	32K	—
Q5_K_M	17.5 GB	✓ Yes	32K	—
Q4_K_M	15.1 GB	✓ Yes	32K	—
Q3_K_M	11.2 GB	✓ Yes	32K	—
Q2_K	9 GB	✓ Yes	32K	—

Assumes a 4K-token context with an f16 KV cache. A longer window needs more; a quantized KV cache needs less. “Max context” is the largest window that still fits in 32 GB. Figures are estimates from parameter count, quantization and memory bandwidth — the analyzer lets you tune KV-cache quant and context.

Context costs VRAM too. At Q4_K_M on the NVIDIA GeForce RTX 5090: 8K 16.1 GB · 32K 21.7 GB. Every window this model publishes still fits 32 GB.

No compatible GPU? Codestral 22B on 16 GB of system RAM, CPU only: runs, but it is tight.