AMD · 16 GB GDDR6 · 640 GB/s bandwidth ·
OpenAI · 21.5B (active 3.6B) · MoE · model ctx up to 128K
Yes — the RX 9070 16GB (16 GB GDDR6) can run gpt-oss-20b. At Q4 with the default 8K context it needs ≈ 13.55 GB VRAM (weights 11.83 GB + KV cache 0.72 GB + 1 GB runtime overhead), which fits within 16 GB. Decode at Q4/8K: ≈ 113.1 tok/s (estimated, batch 1). FP16 needs ≈ 44.72 GB — does not fit on this card. Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.
Fit & speed by quant and context
Quant
Context
VRAM needed
Fits?
tok/s (decode)
Q4
4K
13.19 GB
Yes
113.1est.
Q4
8K default
13.55 GB
Yes
113.1est.
Q4
32K
15.71 GB
No
335.3est.
FP16
4K
44.36 GB
No
11.6est.
FP16
8K default
44.72 GB
No
11.4est.
FP16
32K
46.88 GB
No
10.4est.
Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured
anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s
means the model is far beyond this card (offload-only territory).