Can the Radeon RX 9070 XT Run Llama 3.2 3B?
Runs
Yes. Llama 3.2 3B at Q4_K_M is a 2 GB download (Hugging Face file listing for bartowski/Llama-3.2-3B-Instruct-GGUF), and the Radeon RX 9070 XT has 16 GB of VRAM, leaving about 14 GB for the KV cache and runtime. No published speed measurement for this pair passes SpecPicks' sanity checks yet, so no tokens-per-second figure is quoted.
Buy the Radeon RX 9070 XT
PowerColor Reaper AMD Radeon RX 9070 XT 16GB GDDR6Tracked at $920 on 2026-11-08 — price may vary.
Published runs for Llama 3.2 3B on the Radeon RX 9070 XT
No published run for this pair has passed the checks yet.
More about this pair
How much VRAM does Llama 3.2 3B need?
The Q4_K_M file is 2 GB. Add roughly 1.5 GB for the runtime and an ~8K-token context, so about 3.5 GB of VRAM holds it comfortably. Longer contexts grow the KV cache and need more.
What speed does Llama 3.2 3B reach on the Radeon RX 9070 XT?
No published run of Llama 3.2 3B on the Radeon RX 9070 XT has been cited yet. Generation reads every weight once per token, so the Radeon RX 9070 XT's 645 GB/s of memory bandwidth divided by the 2.02 GB file puts the theoretical ceiling near 319 tokens per second. That is an upper bound, not a speed estimate: kernel, launch and compute overhead keep real runs below it.
Check another pair
Llama 3.2 3B on other GPUs
- RTX 4060 — Runs
- RTX 3060 Ti — Runs
- RTX 3060 12GB — Runs, 126 tok/s median
- Intel Arc B580 — Runs
- RTX 4070 SUPER — Runs
- RTX 4060 Ti 16GB — Runs
- RTX 5060 Ti 16GB — Runs
- Intel Arc A770 16GB — Runs, 64.2 tok/s (1 run)
- RTX 4070 Ti SUPER — Runs
- RTX 4080 — Runs
- RTX 5070 Ti — Runs
- RTX 5080 — Runs, 312 tok/s (1 run)
- Tesla P40 24GB — Runs
- RTX 3090 — Runs
- Radeon RX 7900 XTX — Runs
- RTX 4090 — Runs
- RTX 5090 — Runs
- RTX A6000 48GB — Runs
- RTX PRO 5000 Blackwell 48GB — Runs
- Instinct MI210 64GB — Runs
- RTX PRO 6000 Blackwell 96GB — Runs
Other models on the Radeon RX 9070 XT
- Llama 3.2 1B — Runs
- Mistral 7B — Runs
- Qwen2.5 7B — Runs
- Llama 3.1 8B — Runs
- DeepSeek-R1-Distill-Llama 8B — Runs
- Qwen3 8B — Runs, 90 tok/s median
- Gemma 2 9B — Runs
- Gemma 3 12B — Runs
- Qwen2.5 14B — Runs
- DeepSeek-R1-Distill-Qwen 14B — Runs
- Qwen3 14B — Runs, 50.1 tok/s median
- Phi-4 14B — Runs
- gpt-oss-20b — Runs, 91.9 tok/s (1 run)
- Gemma 3 27B — Offload
- Gemma 2 27B — Offload
- Qwen3 30B-A3B — Offload
- Qwen3 32B — Offload
- Qwen2.5 32B — Offload
- DeepSeek-R1-Distill-Qwen 32B — Offload
- Qwen3 32B (Q6_K) — Offload
- Qwen2.5 32B (Q6_K) — Offload
- DeepSeek-R1-Distill-Qwen 32B (Q6_K) — Offload
- Gemma 3 27B (Q8_0) — Offload
- Llama 3.1 70B — Offload
- Llama 3.3 70B — Offload
- DeepSeek-R1-Distill-Llama 70B — Offload
How the checks work
The fit is the Q4_K_M file size against the card's VRAM, with 1.5 GB kept for the runtime and an ~8K context. A published run counts only if the pair fits, the run is Q4-class, its source does not call it an estimate, it is under the bandwidth ceiling (memory bandwidth ÷ GB read per token: the whole file for a dense model, only the active experts' weights for a mixture-of-experts model such as gpt-oss-20b or Qwen3 30B-A3B, which is why those can post figures above bandwidth ÷ file size), it is not slower than the same card on a model twice its size, a single run is not slower than the same card on a larger dense model, and it is not one figure pasted across many cards. The whole matrix is downloadable as a CSV under CC BY 4.0.
As an Amazon Associate, SpecPicks earns from qualifying purchases. As an eBay Partner Network member, SpecPicks may earn a commission on eBay links.
This page is editorial synthesis based on publicly available information. No independent first-party benchmarking is reported.