Can the Radeon RX 7900 XTX run DeepSeek V4 Flash?
Does not fit at Q4_K_M.
DeepSeek V4 Flash is 284B parameters (MoE; all experts resident). At Q4_K_M the weights are about 158.7 GiB; with the 1.5 GiB runtime reserve that is 160.2 GiB against the card's 24GB, about 136.2 GiB more than the card has.
All three quants
24GB card| Quant | Weights | With reserve | Verdict | Headroom |
|---|---|---|---|---|
| Q4_K_Mthe everyday quant | 158.7 GiB | 160.2 GiB | Does not fit | short 136.2 GiB |
| Q8_0near-lossless | 280.4 GiB | 281.9 GiB | Does not fit | short 257.9 GiB |
| FP16full weights | 529.0 GiB | 530.5 GiB | Does not fit | short 506.5 GiB |
Verdict rules: fits comfortably when weights plus reserve sit within 85% of VRAM; tight when they fit with little left for context; CPU offload when they exceed VRAM by up to 2x (runs, slowly, with layers in system RAM); no beyond that. The KV cache per token depends on the model's architecture and is not modelled here.
Not this card. These fit it comfortably at Q4.
Nothing in the consumer catalog fits DeepSeek V4 Flash at Q4 on a single card; this is multi-GPU or datacenter territory, or a job for hosted inference (see PicksByModel for per-token pricing).
What the Radeon RX 7900 XTX does run well → · DeepSeek V4 Flash on every card →
About the model
DeepSeek V4 Flash by DeepSeek: 1,048,576-token native context, text, released 2026 under the MIT. Model card → · Hosted pricing on PicksByModel →