What can a RX 7900 XTX 24GB run?
24GB · 960 GB/s · 355W · $950
AMD's wide-bandwidth 24 GB: near-3090 speed for the Vulkan faithful, minus the CUDA tax.
Runs great25
Comfortable headroom — daily-driver territory.
| Model | Quant | Est. speed | Max context |
|---|---|---|---|
| deepseek moe 16b chat | FP16 | Infinity tok/s | — |
| qwen2 7b instruct | FP16 | Infinity tok/s | — |
| Qwen3-0.6B | FP16 | 479 tok/s | 33k |
| GPT-OSS-20B | Q5_K_M | 232 tok/s | 48k |
| Gemma-3n-E4B | FP16 | 212 tok/s | 131k |
| SmolLM2-1.7B | FP16 | 175 tok/s | 8k |
| Llama-3.2-3B | FP16 | 93 tok/s | 131k |
| Phi-4-mini | FP16 | 77 tok/s | 131k |
| Gemma-3-4B | FP16 | 77 tok/s | 131k |
| Qwen3-4B | FP16 | 76 tok/s | 33k |
| Gemma-3-12B | Q8_0 | 43 tok/s | 63k |
| Qwen3-14B | Q6_K | 43 tok/s | 33k |
| Mistral-7B-Instruct | FP16 | 41 tok/s | 33k |
| Mistral-Small-24B | Q4_K_M | 41 tok/s | 48k |
| DeepSeek-R1-Distill-Qwen-7B | FP16 | 39 tok/s | 59k |
| Qwen3.8 | MLX 4-bit | 39 tok/s | 44k |
| Qwen3.8 27B | MLX 4-bit | 39 tok/s | 44k |
| Gemma-3-27B | MLX 4-bit | 39 tok/s | 43k |
| Qwen2.5-Coder-7B | FP16 | 39 tok/s | 59k |
| qwen35 27B | MLX 4-bit | 39 tok/s | 42k |
| Phi-4-14B | Q8_0 | 38 tok/s | 16k |
| DeepSeek-R1-Distill-Qwen-14B | Q8_0 | 38 tok/s | 46k |
| Codestral-22B | Q5_K_M | 38 tok/s | 33k |
| Llama-3.1-8B | FP16 | 37 tok/s | 51k |
| Qwen3-8B | FP16 | 37 tok/s | 33k |
Tight squeeze5
It fits, but close the browser tabs and mind the context.
| Model | Quant | Est. speed | Max context |
|---|---|---|---|
| Qwen3-30B-A3B | Q4_K_M | 297 tok/s | 21k |
| Qwen3-Coder-30B-A3B | Q4_K_M | 297 tok/s | 21k |
| DeepSeek-R1-Distill-Qwen-32B | Q4_K_M | 30 tok/s | 12k |
| GLM-4-32B | Q4_K_M | 30 tok/s | 12k |
| Qwen3-32B | Q4_K_M | 30 tok/s | 13k |
Measured on this card0
Everything here was benchmarked on this exact card by someone who owns one — the tables above are our arithmetic; these numbers came off real machines.
Nobody has measured anything on this card yet. Paste a benchmark — it takes one keystroke and no form.