16GB VRAM

Radeon RX 9060 XT 16G — what LLMs can it run?

84 of 145 indexed models fit comfortably in 16GB at 4K context, each at the highest-quality quant that still leaves headroom.

The short answer for Radeon RX 9060 XT 16G

16 GB, 320 GB/s GDDR6. 84 of 145 models in this index fit comfortably at 4K context; 93 load at all.

Biggest that fits
Xing4.0 29B-A4B — 13.6 GB at Q2_K — 2.4 GB spare at 4K, so longer context comes out of a thin margin.Download: bartowski/XingChen-AGI_Xing4.0-29B-A4B-GGUF · XingChen-AGI_Xing4.0-29B-A4B-Q2_K.gguf · 12.84 GB · listing read 2026-11-10
Room to grow
Stable LM 2 12B Chat — 8.3 GB at Q4_K_M — under 60% of the card, which leaves 7.7 GB for a long context window or a second process.
Speed ceiling
Llama 3.1 8B Instruct at Q4_K_M reads 4.6 GB of weights per token, so 320 GB/s puts a hard ceiling near 69 tok/s. That is arithmetic on two published numbers, not a benchmark — real throughput lands below it. No run on this card has been measured here, so there is nothing to compare the ceiling against.
The wall you will hit
Backend coverage, not capacity. At 320 GB/s the memory system is competitive with NVIDIA cards of the same size; what varies is whether your runtime has a working ROCm build for your kernel and card. llama.cpp and vLLM both ship official ROCm support — this is a setup question, not a hardware ceiling.
Best local LLM for 16GB

32B · 8

Radeon RX 9060 XT 16G — what LLMs can it run? — 32B
ModelQuantEst. VRAMHeadroomtok/s
Xing4.0 29B-A4B29B-A4BQ2_K13.61 GB+2.4 GB≤205Ceiling
Hy-Memory-30B-A3B30B-A3BQ8_0-expsIQ2_XXS9.95 GB+6.1 GB≤370Ceiling
Arete 27B27BQ3_K_M14.03 GB+2 GB19–26Derived
Mistral Small 24B Instruct24BAWQ INT413.23 GB+2.8 GB≤28Ceiling
Devstral Small 1.1 24B24BAWQ INT413.01 GB+3 GB≤29Ceiling
Magistral Small 1.2 24B24BAWQ INT413.01 GB+3 GB≤29Ceiling
Codestral 22B22BAWQ INT412.56 GB+3.4 GB≤30Ceiling
GPT-OSS 20B21B MoEMXFP412.76 GB+3.2 GB≤162Ceiling

14B · 22

Radeon RX 9060 XT 16G — what LLMs can it run? — 14B
ModelQuantEst. VRAMHeadroomtok/s
InternLM2 20B Chat20BQ4_K_M13.43 GB+2.6 GB21–28Derived
DeepSeek-Coder-V2-Lite Instruct16B-A2.4BQ4_K_M10.08 GB+5.9 GB≤232Ceiling
DeepSeek-V2-Lite Chat16B-A2.4BQ4_K_M10.08 GB+5.9 GB≤232Ceiling
StarCoder2 15B15BQ4_K_M10.19 GB+5.8 GB28–37Derived
Qwen3 14B Instruct14BQ5_K_M11.65 GB+4.4 GB24–32Derived
Qwen2.5 14B Instruct14BQ5_K_M11.73 GB+4.3 GB25–32Derived
DeepSeek-R1-Distill-Qwen-14B14BQ4_K_M10.14 GB+5.9 GB29–38Derived
Phi-4 14B14BQ5_K_M11.77 GB+4.2 GB25–32Derived
Phi-3 Medium 14B Instruct14BQ6_K12.86 GB+3.1 GB22–29Derived
Ministral 3 14B Instruct14BQ5_K_M10.74 GB+5.3 GB27–35Derived
Mistral Nemo 12B Instruct12BQ6_K11.14 GB+4.9 GB26–34Derived
Gemma 3 12B IT12BQ5_K_M9.84 GB+6.2 GB30–39Derived
Mellum2.1 12B-A2.5B Thinking12B-A2.5BQ8_013.6 GB+2.4 GB≤127Ceiling
Stable LM 2 12B Chat12BQ4_K_M8.35 GB+7.7 GB35–46Derived
Gemma 4 12B IT12BQ8_013.82 GB+2.2 GB20–27Derived
Llama 3.2 11B Vision Instruct11BQ8_012.9 GB+3.1 GB22–29Derived
Solar 10.7B Instruct11BQ4_K_M7.6 GB+8.4 GB39–52Derived
Falcon 3 10B Instruct10BQ4_K_M7.28 GB+8.7 GB41–53Derived
Metis-1.6 Logos10B MoEQ8_011.67 GB+4.3 GB≤121Ceiling
Gemma 2 9B Instruct9BQ8_011.7 GB+4.3 GB26–34Derived
Qwen3.5 9B9BQ8_010.39 GB+5.6 GB26–34Derived
K2 Horizon 7B7BQ8_010.62 GB+5.4 GB27–35Derived

7B · 31

Radeon RX 9060 XT 16G — what LLMs can it run? — 7B
ModelQuantEst. VRAMHeadroomtok/s
GLM-4-9B-Chat9BQ8_010.16 GB+5.8 GB27–35Derived
MiMo-V2.6-Distill-Qwen-9B9BQ8_010.11 GB+5.9 GB27–35Derived
Qwen3-VL 8B Instruct8BQ8_010.39 GB+5.6 GB27–36Derived
TwIL 8B balanced8.8BQ4_K_M6.28 GB+9.7 GB48–63Derived
Granite 4.2 8B8BQ8_010.46 GB+5.5 GB27–36Derived
Ministral 3 8B Instruct8BQ8_010.02 GB+6 GB28–37Derived
LFM2.5 8B-A1B8B-A1.5BQ8_09.46 GB+6.5 GB≤211Ceiling
Qwen2-VL 7B Instruct7BQ4_K_M5.49 GB+10.5 GB51–67Derived
Granite 3.1 8B Instruct8BQ4_K_M5.88 GB+10.1 GB52–68Derived
Qwen3 8B Instruct8BQ6_K7.64 GB+8.4 GB38–50Derived
Llama 3.1 8B Instruct8BQ8_09.47 GB+6.5 GB30–40Derived
Nous Hermes 3 Llama 3.1 8B8BQ4_K_M5.64 GB+10.4 GB53–69Derived
Aya 23 8B8BQ4_K_M5.64 GB+10.4 GB53–69Derived
OpenChat 3.6 8B8BQ4_K_M5.64 GB+10.4 GB53–69Derived
DeepSeek-R1-Distill-Llama-8B8BQ5_K_M6.51 GB+9.5 GB45–59Derived
InternLM2 7B Chat7BQ4_K_M5.45 GB+10.6 GB55–72Derived
Qwen2.5 7B Instruct7BQ6_K6.77 GB+9.2 GB41–54Derived
Qwen2.5-Coder 7B Instruct7BQ4_K_M5.07 GB+10.9 GB55–73Derived
DeepSeek-R1-Distill-Qwen-7B7BQ4_K_M5.07 GB+10.9 GB55–73Derived
Gemma 4 E4B ITE4BQ8_08.68 GB+7.3 GB31–41Derived
Nemotron 3.5 Bolt E4BE4BQ8_08.5 GB+7.5 GB32–42Derived
OLMo 2 7B Instruct7BQ8_010.3 GB+5.7 GB33–43Derived
OLMo 3 7B Instruct7BQ8_010.3 GB+5.7 GB33–43Derived
WizardLM-2 7B7BQ4_K_M5.14 GB+10.9 GB58–77Derived
Mistral 7B Instruct v0.37BQ6_K6.75 GB+9.3 GB43–57Derived
Zephyr 7B Beta7BQ6_K6.75 GB+9.3 GB43–57Derived
K2 Horizon 3.7B3.7BQ8_06.24 GB+9.8 GB≤63Ceiling
Gemma 4 E2B ITE2BQ8_05.2 GB+10.8 GB≤68Ceiling
Qwen3.5 4B4BQ8_04.97 GB+11 GB≤73Ceiling
Gemma 3 4B IT4BQ8_05.05 GB+11 GB≤74Ceiling
Spark-X2.6 4B4BQ8_04.84 GB+11.2 GB≤77Ceiling

≤3B · 23

Radeon RX 9060 XT 16G — what LLMs can it run? — ≤3B
ModelQuantEst. VRAMHeadroomtok/s
Qwen3 4B Instruct4BQ6_K4.05 GB+12 GB≤103Ceiling
Phi-4 Mini Instruct3.8BQ8_04.81 GB+11.2 GB≤82Ceiling
Phi-3.5 Mini Instruct3.8BQ8_05.89 GB+10.1 GB≤83Ceiling
Granite 4.2 3B3BQ8_04.41 GB+11.6 GB≤87Ceiling
Ministral 3 3B Instruct3BQ8_04.26 GB+11.7 GB≤92Ceiling
Llama 3.2 3B Instruct3BQ8_04.05 GB+12 GB≤99Ceiling
Qwen2.5 3B Instruct3BQ8_03.59 GB+12.4 GB≤103Ceiling
SmolLM3 3B3BQ8_03.73 GB+12.3 GB≤103Ceiling
Metis-1.6 Praxis3B MoEQ8_03.4 GB+12.6 GB≤283Ceiling
Gemma 2 2B Instruct2BQ8_03.34 GB+12.7 GB≤122Ceiling
Qwen3.5 2B2BQ8_02.23 GB+13.8 GB≤162Ceiling
LFM3 1.7B1.7BQ8_01.99 GB+14 GB≤183Ceiling
LFM3-VL 2B2BQ8_01.99 GB+14 GB≤183Ceiling
Qwen3 1.7B Instruct1.7BQ8_02.39 GB+13.6 GB≤184Ceiling
Qwen2.5 1.5B Instruct1.5BQ8_01.83 GB+14.2 GB≤206Ceiling
Llama 3.2 1B Instruct1BQ8_01.51 GB+14.5 GB≤256Ceiling
LFM2.5 1.2B Instruct1.2BQ8_01.35 GB+14.7 GB≤271Ceiling
K2 Horizon 0.9B0.9BQ8_01.44 GB+14.6 GB≤294Ceiling
Qwen3.5 0.8B0.8BQ8_00.94 GB+15.1 GB≤395Ceiling
BotPsy 0.8B0.75BQ8_00.9 GB+15.1 GB≤416Ceiling
Qwen2.5 0.5B Instruct0.5BQ8_00.6 GB+15.4 GB≤653Ceiling
MNN-Pocket 0.3B Instruct0.36BQ8_00.46 GB+15.5 GB≤865Ceiling
MNN-Pocket 0.3B QAT0.30BQ4_10.26 GB+15.7 GB≤1778Ceiling

How this list is built

Each row is the lowest-perplexity-loss quant of that model whose estimated total — weights plus KV cache at 4K context plus activation buffer — uses at most 88% of the card. That is the calculator's "green" threshold, so every row here has real headroom rather than only just fitting. Raise the context length and the list shortens; the calculator lets you check any combination directly. The speed column is this card: a measured run, a range scaled from this card's own Llama 2 7B benchmark, or a bandwidth ceiling. A figure that only exists for the RTX 4090 is labelled as that card.

A further 9 models load but with no headroom to spare (up to 105% of VRAM) — the Quant Hub’s GPU chips count those too, which is why its number is higher.

Measured on this card

No benchmark runs in this index were recorded on a Radeon RX 9060 XT 16G. Every figure on this page is calculated from the model architecture and the quant level — treat them as estimates, not measurements.

Cards with the same budget

What fits is decided by memory, so every 16GB card of this type returns the same list. These pages are not different answers — they differ in throughput, which this index does not measure per card.

The same 16GB in system memory is a different machine — everything on this page fits there, at a speed set by your DIMMs rather than a graphics bus: 16 GB RAM (CPU)

Same list, different speed — play one sentence on both cards: Radeon RX 9060 XT 16G vs Radeon RX 9070 XT (Arete 27B) · Radeon RX 9060 XT 16G vs Radeon RX 9070 (Arete 27B) · Radeon RX 9060 XT 16G vs Radeon RX 7900 GRE (Arete 27B)

Setup guides for this card

Step-by-step guides written for the Radeon RX 9060 XT 16G's runtime or memory size, with the commands and the checks that prove the model is running on the card.

Stepping up

A Radeon RX 7900 XT (20GB) fits 12 more of the indexed models than this card. Radeon RX 7900 XT →

Before buying, hear the speed difference on a model both cards hold: Radeon RX 9060 XT 16G vs Radeon RX 7900 XT (Arete 27B) →

Common questions

What is the best local LLM for a Radeon RX 9060 XT 16G?

For everyday use, Stable LM 2 12B Chat at Q4_K_M — about 8.3 GB of the card's 16 GB at 4K context, leaving 7.7 GB for a longer window. If you want the largest thing that will load, that is Xing4.0 29B-A4B at Q2_K (13.6 GB, 2.4 GB spare). "Best" here means best fit for the memory budget — this index does not run task benchmarks, so it cannot tell you which model is smarter.

Can a Radeon RX 9060 XT 16G run Nemotron 3.5 Lightning 30B-A3B?

No. Its smallest build here, Q4_0, needs about 19.8 GB and this card has 16.0 GB usable — short by 3.8 GB before any context beyond 4K. The largest model this card does clear is Xing4.0 29B-A4B.

How many tokens per second does a Radeon RX 9060 XT 16G do on an 8B model at Q4?

This index has no measured run on this card, so it does not publish a figure. What can be stated from specifications: generating a token requires reading every weight once, Llama 3.1 8B Instruct at Q4_K_M is 4.6 GB of weights, and this card moves 320 GB/s — a ceiling near 69 tok/s. Batch size, context length, the runtime and how much of the model sits in cache all take you below it.

Radeon RX 9060 XT 16G or RTX 5080 for local LLMs?

They hold the same models: 16 GB against 16 GB fits 84 and 84 of 145 respectively at 4K. The difference is throughput — 320 GB/s against 960 GB/s, a 3.0× gap in how fast the weights can be read, which is what token generation is bound by. The RTX 5080 generates faster on any model both can hold.

84 of 145 indexed models fit comfortably in 16GB at 4K context, each at the highest-quality quant that still leaves headroom.

This page's figures change when the model or the runtime does.Last updated 2026-11-17 RSS → /feed.xml