HARDWARE / VRAM

GeForce RTX 4090 · 24 GB

A 27B Q4 checkpoint can be a tight fit. Leave room for runtime overhead and KV cache instead of filling every gigabyte with weights.

Birincil kaynak

Consumer Ada GPU24 GB

01 / PRACTICAL CEILING

Qwen3 30B-A3B

The highest listed result may be a tight or offloaded fit. For daily use, prefer the first result marked Comfortable and keep context modest.

02 / DEFAULT RUNTIME

llama.cpp

GGUF models, CPU/GPU offload, embedded and unusual hardware. New architectures may require a recent build. Backend availability does not imply equal performance.

Rehberi aç

03 / MODEL ENVELOPE

Temsilî açık ağırlıklı modeller

Alibaba Qwen

Qwen3 30B-A3B

30B3B active · GGUF Q4 MoE

Tahmini bellek: ~22.5 GB
Bağlam: 32K+

Alibaba Qwen

Qwen3.6 27B

27BQ4 / NVFP4 estimate

Tahmini bellek: ~18.5 GB
Bağlam: 262K native

Alibaba Qwen

Qwen3.8 27B

27BQ4 estimate

Tahmini bellek: ~19.5 GB
Bağlam: 262K native

Google

Gemma 3 27B

27BINT4

Tahmini bellek: ~20.5 GB
Bağlam: 128K

Mistral AI

Devstral Small 2 24B

24BGGUF Q4 / BF16

Tahmini bellek: ~18.5 GB
Bağlam: 256K

OpenAI

gpt-oss-20b

21B3.6B active · MXFP4

Tahmini bellek: ~16 GB
Bağlam: 128K

Dikkat

Fit is based on estimated total model memory. Driver support, KV cache, multimodal projectors, concurrency, and desktop applications can all reduce available headroom. Current fit: Qwen3 30B-A3B (Tight), Qwen3.6 27B (Tight), Qwen3.8 27B (Tight), Gemma 3 27B (Tight), Devstral Small 2 24B (Tight), gpt-oss-20b (Comfortable).