Can the Radeon RX 7900 XTX run Qwen3.6-35B-A3B?
Yes. A new Radeon RX 7900 XTX ($1,350) runs Qwen3.6-35B-A3B at Q4 with room for about 33K tokens of context (63K with 8-bit KV cache), at about 137 tok/s estimate. That is $10 per tok/s, #2 of 5 priced GPUs that fit it; the GeForce RTX 3090 ranks first, also at $10.
Yes, with room for long context. On the AMD Radeon RX 7900 XTX (24 GB) there is room for about 33K tokens of context with the default FP16 KV cache, or 63K with an 8-bit (q8_0) KV cacheestimate: what is left after the weights and a 1 GB runtime reserve. That is 32K tokens or more, enough for long documents and code. The Q4_K_M GGUF of Qwen3.6-35B-A3B is 22.29 GB (source); with our 1.5 GB allowance for the runtime and a short context it needs about 23.8 GB.
Qwen3.6-35B-A3B fits 5 of the 18 GPUs we have a dated price for; the smallest of them has 24 GB. The Radeon RX 7900 XTX's 960 GB/s gives ~137 tok/s estimate, 72% of the fastest card, the GeForce RTX 5090 at ~189 tok/s. Only 3B of its 35.95B parameters are active per token, so the estimate reads 1.86 GB per token rather than the whole 22.29 GB file, plus a per-token routing and expert-gather cost calibrated to 45 measured llama.cpp benchmarks (2025–26). On the same card, the mixture-of-experts K2-Horizon-MoVA-36B-A4B (4B of 37.44B active, 22.37 GB) estimates ~178 tok/s.
At $1,350 (new) it works out to $10 per tok/s, #2 of 5. The GeForce RTX 3090 is the cheapest per tok/s at $10 ($1,499, used). The cheapest complete build we priced that runs Qwen3.6-35B-A3B is our 24 GB: one used RTX 3090 build at $2,650.58.
Last updated Oct 16, 2026. Speed is a bandwidth-based estimate (960 GB/s, AMD), scaled by per-platform constants calibrated to 45 measured llama.cpp benchmarks (2025–26), not a benchmark of this pair. Max context is an estimate from the model's architecture for llama.cpp with flash attention.
Next size up: NVIDIA Nemotron 3.5 Lightning 30B-A3B (25.48 GB at Q4_K_M) needs ~27 GB and does not fit on one AMD Radeon RX 7900 XTX. See what runs NVIDIA Nemotron 3.5 Lightning 30B-A3B.
Every priced GPU that runs Qwen3.6-35B-A3B, by cost per token/sec
| GPU | VRAM | Speed estimate | Max context estimate | Low price | $ per tok/s |
|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3090 | 24 GB | ~157 tok/s | 33K · 63K (8-bit) | $1,499 (used) | $10 |
| AMD Radeon RX 7900 XTX | 24 GB | ~137 tok/s | 33K · 63K (8-bit) | $1,350 (new) | $10 |
| NVIDIA GeForce RTX 4090 | 24 GB | ~161 tok/s | 33K · 63K (8-bit) | $2,910 (used) | $18 |
| NVIDIA GeForce RTX 5090 | 32 GB | ~189 tok/s | 256K (full) | $6,500 (new) | $34 |
| NVIDIA RTX A6000 | 48 GB | ~146 tok/s | 256K (full) | $8,363 (new) | $57 |
Prices: picksbycard.com, 2026-11-02; see how prices are sourced. Prices move daily. Bar length shows value: longer is cheaper per tok/s.