Radeon RX 6950 XT Price & Used Prices: Which Local LLMs Can It Run?

AMD's RX 6950 XT (2022) is a 16GB gaming flagship running Gemma 4 12B at 4-bit at about 56 tokens/s. ROCm doesn't list it; on Windows, use Vulkan.

eBay

Radeon RX 6950 XT price history

$680-1.73%

eBay

Price updated 10/23/2026, 09:52

$1,102$923$744$565$386

Low $580High $1,000

Sep 8Sep 30Oct 23

As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.

Full Radeon RX 6950 XT specifications

Memory16 GB · GDDR6
Bandwidth576 GB/s
FP16 / BF1647.3 TFLOPS
INT894.6 TOPS

Bus width: 256-bit
Architecture: RDNA 2 (Navi 21)
Cores: 5,120
Ecosystem: ROCm
Interface: PCIe 4.0 x16
Power: 335 W
Power connector: 2× 8-pin
Type: Consumer
Launch date: May 10, 2022
MSRP: $1,099

Best Radeon RX 6950 XT deals on eBay

Deals from trusted eBay sellers, picked by our rules. Shop with confidence.

Updated 1 hour ago1 hr. ago

As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.

See more on eBay

What LLMs can a single Radeon RX 6950 XT run?

One 6950 XT is steadiest with up to 12B at 4-bit. 30B-class MoE at 2-bit is fastest, about 130 tokens/s, though long documents take a while to load.

Model	Size	Runs?	Speed t/s	Max context
Gemma 4 E4B8B	3.8 GB	Runs	98.2	128K
Qwen3.5 9B9.7B	4.1 GB	Runs	91	256K
Gemma 4 12B12B	4.7 GB	Runs	79.6	256K
Gemma 4 26B A4B25.2B (3.8B active)	10.5 GB	Runs	114.9	256K
Qwen3.6 27B27.8B	11.8 GB	Runs	34.8	64K
Qwen3.8 27B27.8B	9.8 GB	Runs	41.3	93K
GLM 4.7 Flash30B (3B active)	11.9 GB	Runs	126.8	75K
Gemma 4 31B30.7B	11.8 GB	Runs	33.6	41K
Qwen3.6 35B A3B36B (3B active)	12.3 GB	Runs	138.7	181K

Model	Size	Runs?	Speed t/s	Max context
Gemma 4 E4B8B	5 GB	Runs	77.5	128K
Qwen3.5 9B9.7B	5.7 GB	Runs	68.5	256K
Gemma 4 12B12B	7.1 GB	Runs	55.6	256K
Gemma 4 26B A4B25.2B (3.8B active)	16.9 GB	Needs RAM offload	36.6	256K
Qwen3.6 27B27.8B	16.8 GB	Too big	—	—
Qwen3.8 27B27.8B	16.5 GB	Too big	—	—
GLM 4.7 Flash30B (3B active)	18.3 GB	Needs RAM offload	35.5	198K
Gemma 4 31B30.7B	18.3 GB	Too big	—	—
Qwen3.6 35B A3B36B (3B active)	22.1 GB	Needs RAM offload	49.3	256K

Model	Size	Runs?	Speed t/s	Max context
Gemma 4 E4B8B	8.2 GB	Runs	49.6	128K
Qwen3.5 9B9.7B	9.5 GB	Runs	43.1	196K
Gemma 4 12B12B	12.7 GB	Runs	32.6	183K
Gemma 4 26B A4B25.2B (3.8B active)	26.9 GB	Needs RAM offload	24.7	256K
Qwen3.6 27B27.8B	28.6 GB	Too big	—	—
Qwen3.8 27B27.8B	29 GB	Too big	—	—
GLM 4.7 Flash30B (3B active)	31.8 GB	Needs RAM offload	22.1	198K
Gemma 4 31B30.7B	32.6 GB	Too big	—	—
Qwen3.6 35B A3B36B (3B active)	36.9 GB	Needs RAM offload	32.7	256K

RunsFits fully in VRAM; speed is estimated from VRAM bandwidth.

Needs RAM offloadMoE only: expert weights sit in system RAM, so speed is estimated from 70 GB/s RAM bandwidth.

Too bigWon’t fit on one card at this quantization, even with RAM offload.

Full calculation rules are on the Methodology page.

Who it suits

Good fit

You already own this card and want to run models up to 12B
Gemma 4 12B and Qwen3.5 9B at 4-bit run at about 56 to 69 tokens/s
Mainly using llama.cpp or Ollama on Linux
It shares a chip with the officially supported PRO W6800, so ROCm recognizes it directly, with no environment variable tweaks
You can live with 2-bit
Qwen3.6 35B A3B at 2-bit needs only 12.3 GB and runs entirely on the card at about 139 tokens/s

Not a good fit

Running 27B dense models at 4-bit
Qwen3.8 27B needs 16.5 GB, more than 16GB, so you have to drop to 2-bit
Your main system runs Windows
Neither AMD's validated llama.cpp prebuilt packages nor Ollama's Windows ROCm support list includes this card
You need vLLM
This generation isn't among its officially supported AMD cards

Before you buy

Decide between Linux and Windows first
Linux is mostly fine: it shares a chip with the officially supported PRO W6800, ROCm libraries are already compiled for it, and Ollama's Linux support list includes the 6950 XT. It just isn't on AMD's official support list, so if something breaks, there is no official backing. Windows is harder: AMD's validated llama.cpp prebuilt packages only cover newer cards, and in July 2026 Ollama removed it from its Windows ROCm list, supporting it through the Vulkan backend instead.
No matrix units, so long prompts are slow
This generation has no matrix compute units, and its FP16 compute is only a bit over 60% of the RX 7800 XT's. Generation speed depends on bandwidth, so it is barely affected. But for uses that process upward of ten thousand tokens up front every time, such as reading long documents, RAG or coding assistants, the wait for the first token is noticeably longer. It also lacks BF16 support, which doesn't matter for GGUF quantized models.
Power supply
The card draws 335W through two 8-pin connectors, and AMD recommends an 850W system power supply. The RX 7800 XT, also 16GB, needs only 263W with a recommended 700W supply; if you leave models running for long stretches, that 70W difference is there the whole time.

Compared with similar GPUs

Radeon RX 6950 XT vs Radeon RX 7800 XT, GeForce RTX 5060 Ti 16GB, and Radeon RX 7900 XTX: price, VRAM, bandwidth, compute and local LLM speed compared.

Spec	Radeon RX 6950 XT	Radeon RX 7800 XT	GeForce RTX 5060 Ti 16GB	Radeon RX 7900 XTX
Memory	16 GB	16 GBSame	16 GBSame	24 GB50% more
Bandwidth	576 GB/s	624 GB/s8% higher	448 GB/s22% lower	960 GB/s67% higher
BF16 / FP16 compute	47 TFLOPS	75 TFLOPS58% higher	95 TFLOPS101% higher	123 TFLOPS160% higher
INT8 compute	95 TOPS	75 TOPS21% lower	190 TOPS101% higher	123 TOPS30% higher
Qwen3.8 27B speed (2-bit)	41.3 t/s	44.6 t/s8% faster	32.6 t/s21% slower	66.3 t/s61% faster
Largest dense model (4-bit)	23B	23BSame	23BSame	37B61% more
Power	335 W	263 W21% lower	180 W46% lower	355 W6% higher
Interface	PCIe 4.0 x16	PCIe 4.0 x16	PCIe 5.0 x8	PCIe 4.0 x16
Released	2022-05	2023-09	2025-04	2022-12
eBay median	$680Buy Radeon RX 6950 XT	$649Buy Radeon RX 7800 XT	$845Buy GeForce RTX 5060 Ti 16GB	$1,250Buy Radeon RX 7900 XTX