Best AI Models for AMD Radeon RX 6950 XT (16 GB)

Runs 14B models well; 27–31B only just fits.

16 GB GDDR6576 GB/s47 TFLOPS FP16335 WRDNA 2 2022

16 GB fits ~14B models at Q4-Q6 or 24B-class at low quants. Ollama supports it on Linux; on Windows use the Vulkan backend (llama.cpp, LM Studio).

What LLMs can the RX 6950 XT run?

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 106 t/s	106 tok/s	Q4_K_M · 3.4 GB
3	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 81.3 t/s	81.3 tok/s	Q4_K_M · 13.1 GB
4	Phi 4Microsoft Q4_K_M · 12.5 GB · 36.3 t/s	36.3 tok/s	Q4_K_M · 12.5 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 42.0 t/s	42.0 tok/s	Q4_K_M · 9.0 GB

#	Model	Speed	Coding Score	Quant · VRAM
1	GPT OSS 20BOpenAIHow to run → Q4_K_M · 13.1 GB · 81.3 t/s	81.3 tok/s		Q4_K_M · 13.1 GB
2	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 65.2 t/s	65.2 tok/s		Q4_K_M · 7.9 GB
3	Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	overall	Q4_K_M · 6.6 GB
4	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 42.0 t/s	42.0 tok/s		Q4_K_M · 9.0 GB
5	Ling 3.0 TinyInclusion AI Q4_K_M · 5.5 GB · 133 t/s	133 tok/s		Q4_K_M · 5.5 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	Q4_K_M · 6.6 GB
2	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 81.3 t/s	81.3 tok/s	Q4_K_M · 13.1 GB
3	Qwen3 8BAlibaba Q4_K_M · 7.6 GB · 62.6 t/s	62.6 tok/s	Q4_K_M · 7.6 GB
4	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 65.2 t/s	65.2 tok/s	Q4_K_M · 7.9 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 42.0 t/s	42.0 tok/s	Q4_K_M · 9.0 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 106 t/s	106 tok/s	Q4_K_M · 3.4 GB
3	Mistral Nemo Instruct 2407Mistral AI Q4_K_M · 11.0 GB · 42.8 t/s	42.8 tok/s	Q4_K_M · 11.0 GB
4	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 65.2 t/s	65.2 tok/s	Q4_K_M · 7.9 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 42.0 t/s	42.0 tok/s	Q4_K_M · 9.0 GB

#	Model	Speed ↓	Quant · VRAM
1	Qwen3.5 4BAlibabaHow to run → Q4_K_M · 3.4 GB · 106 t/s	106 tok/s	Q4_K_M · 3.4 GB
2	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 81.3 t/s	81.3 tok/s	Q4_K_M · 13.1 GB
3	Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	Q4_K_M · 6.6 GB
4	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 42.0 t/s	42.0 tok/s	Q4_K_M · 9.0 GB
5	Phi 4Microsoft Q4_K_M · 12.5 GB · 36.3 t/s	36.3 tok/s	Q4_K_M · 12.5 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 54.3 t/s	54.3 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 106 t/s	106 tok/s	Q4_K_M · 3.4 GB
3	Gemma 3 12B ITGoogle Q4_K_M · 10.9 GB · 43.0 t/s	43.0 tok/s	Q4_K_M · 10.9 GB
4	Gemma 3 4B ITGoogle Q4_K_M · 5.0 GB · 122 t/s	122 tok/s	Q4_K_M · 5.0 GB
5	MiMo V2.6 Distill Qwen 9BXiaomi Q4_K_M · 6.5 GB · 55.7 t/s	55.7 tok/s	Q4_K_M · 6.5 GB

See all 309 models on the RX 6950 XT →Best quant that fits · speeds estimated on this card

About the RX 6950 XT

The RX 6950 XT is a solid mid-range card in the 16 GB class: it runs 14B models well. It is power-hungry for its speed, and among the best speed per dollar of the cards compared at ≈$550 used. RDNA 2 generation, released 2022.

Speed in 16 GB#9576 GB/s

Per watt#150.16 tok/s per W on Qwen3.5 9B (9B params)

Per dollar#29.9 tok/s per $100 on Qwen3.5 9B (9B params)

GenerationRDNA 2released 2022

RX 6950 XT strengths and weaknesses

Strengths

Runs 35B MoE models: Qwen3.6 35B A3B at 71.5 tok/s in 15.7 GB

Weaknesses

70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 16 GB here
Slow prompt reading: A 4K-token prompt on Qwen3.5 9B waits 7.6 s for the first token
ROCm, not CUDA: No FP8 weights; MXFP4 only through llama.cpp, not vLLM

Speed per dollar and per watt

NewUsed

On Qwen3.5 9B (9B params) at Q4_K_M. New prices are compared with new, used with used. The last two columns are the speed you get per $100 of price and per 100 W of board power; higher is better. The RX 6950 XT has no new price, so the New view shows it at its MSRP.

Card	New price	tok/s	tok/s per $100	tok/s per 100 W
RTX 507012 GB	about $840	68.7	8.2	27.5
RTX 5070 Ti16 GBBest new value	about $1,170	91.6	7.8	30.5
RX 907016 GB	about $800	60.4	7.6	27.5
RX 9070 XT16 GB	about $815	60.4	7.4	19.9
RX 9060 XT 16GB16 GB	about $590	30.2	5.1	18.9
RX 6950 XT16 GB	about $1,099	54.3	4.9	16.2
RTX 508016 GB	about $2,394	98.1	4.1	27.3
RTX 509032 GB	about $5,502	183	3.3	31.8

Card	Used price	tok/s	tok/s per $100	tok/s per 100 W
RTX 3060 12GB12 GBBest used value	about $329	36.8	11.2	21.6
RX 6950 XT16 GB	about $550	54.3	9.9	16.2
RX 907016 GB	about $650	60.4	9.3	27.5
RTX 407012 GB	about $575	51.5	9.0	25.8
RTX 5070 Ti16 GB	about $1,100	91.6	8.3	30.5
RTX 4070 SUPER12 GB	about $675	51.5	7.6	23.4
RX 7900 XTX24 GB	about $1,200	90.6	7.6	25.5
RX 9060 XT 16GB16 GB	about $550	30.2	5.5	18.9

Similar GPUs

Cards people weigh against the RX 6950 XT, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.

11% faster, same 16 GBRX 9070AMD · 16 GB · RDNA 4

▲ 11% speedsame 16 GB$800

69% faster, same 16 GBRTX 5070 TiNVIDIA · 16 GB · Blackwell

▲ 69% speedsame 16 GB$1,170rent $0.20/hr

Same 16 GB, new with warrantyRX 9060 XT 16GBAMD · 16 GB · RDNA 4

▼ 44% speedsame 16 GB$590

8 GB more and 67% fasterRX 7900 XTXAMD · 24 GB · RDNA 3

▲ 67% speed▲ 8 GB$999

32% slower, 12 GBRTX 3060 12GBNVIDIA · 12 GB · Ampere

▼ 32% speed▼ 4 GB$329rent $0.06/hr

More efficient per wattRX 7900 XTAMD · 20 GB · RDNA 3

▲ 39% speed▲ 4 GB$899

RX 6950 XT tokens per second by model size

Output speed depends on memory bandwidth (576 GB/s). Grey rows need more than 16 GB.

9BQwen3.5 9B · Q4_K_M · 6.6 GB
54.3tok/s
14BQwen3 14B · Q4_K_M · 11.9 GB
36.4tok/s
27–31BQwen3.8 27B · Q3_K_L · 15.4 GB tight · Q3_K_L
23.1tok/s
35B MoEQwen3.6 35B A3B · IQ3_S · 15.7 GB tight · IQ3_S
71.5tok/s
70BLlama 3.3 70B Instruct · not at 4-bit
needs 51.3 GB
120B MoEGPT OSS 120B · not at 4-bit
needs 70.8 GB

Estimated speeds: Q4_K_M, 4K context, one user. How we estimate

All 1073 models on the RX 6950 XT

Featured only

Best quant that fits, ordered by llmrun Score. Fit grades how well it runs here, S best. How ranking works

LLM models compatible with AMD Radeon RX 6950 XT — ranked by Score, recency and Fit
Model	Quant	VRAM	Speed	Context	Fit
Qwen3 30B A3B.w8a830.6BChatFunctions Q3_K_M·69.0 t/s	Q3_K_M	16 GB100%	69.0 t/s	41K	C44, Usable
Mistral Small 3.2 24B Instruct 250624.0BChat Q3_K_L·26.5 t/s	Q3_K_L	16.0 GB100%	26.5 t/s	131K	C35, Usable
Falcon H1 7B Base7.6BChat Q4_K_M·68.6 t/s	Q4_K_M	6.3 GB40%	68.6 t/s	262K	S85, Runs great
Tiny BloomForCausalLM2MChat Q4_K_M·34560.0 t/s	Q4_K_M	0.3 GB2%	34560.0 t/s	—	S100, Runs great
Qwen3 0.6B Base596MChat Q4_K_M·443.1 t/s	Q4_K_M	1.6 GB10%	443.1 t/s	33K	S100, Runs great
MiniCPM V 2 68.1BVision Q4_K_M·64.6 t/s	Q4_K_M	7.9 GB50%	64.6 t/s	—	A84, Runs well
Huihui Qwen3 Coder 30B A3B Instruct Abliterated30.5BChatCodeFunctions Q3_K_M·69.2 t/s	Q3_K_M	16.0 GB100%	69.2 t/s	262K	C44, Usable
Mistral Small 24B Instruct 2501 Quantized.w8a823.6BChat Q3_K_M·28.3 t/s	Q3_K_M	15.2 GB95%	28.3 t/s	33K	B46, Decent
Granite 3.2 8B Instruct8.2BChatFunctions Q4_K_M·62.4 t/s	Q4_K_M	7.9 GB49%	62.4 t/s	131K	A84, Runs well
Browsesafe30.5BChat Q3_K_M·69.2 t/s	Q3_K_M	16.0 GB100%	69.2 t/s	262K	C44, Usable
Mistral Small 3.2 24B Qiskit24.0BChatCode Q3_K_M·27.8 t/s	Q3_K_M	15.4 GB96%	27.8 t/s	131K	C44, Usable
NVIDIA Nemotron 3.5 Lightning 30B A3B Base BF1631.6BChat Q3_K_M·72.4 t/s	Q3_K_M	15.8 GB99%	72.4 t/s	262K	C44, Usable
Mellum2 12B A2.5B Instruct SFT12.1BChat Q4_K_M·99.5 t/s	Q4_K_M	7.8 GB48%	99.5 t/s	131K	S91, Runs great
Qwen2.5 7B Instruct 1M7.6BChatFunctions Q4_K_M·69.3 t/s	Q4_K_M	5.8 GB36%	69.3 t/s	1010K	S86, Runs great
Medgemma 4B IT4.3BVision Q4_K_M·121.7 t/s	Q4_K_M	5 GB31%	121.7 t/s	—	S95, Runs great

RX 6950 XT specs, power and PSU

Specs

Architecture: RDNA 2
VRAM: 16 GB GDDR6
Memory bandwidth: 576.0 GB/s
FP16 (dense): 47.31 TFLOPS
Stream processors: 5,120
TDP: 335 W
Released: 2022-05-10
MSRP: $1,099

RX 6950 XT FAQ

Is the RX 6950 XT still worth it in 2026?

At ≈$550 used you get 9.9 tok/s per $100 on Qwen3.5 9B, which ranks 2nd of the 8 cards compared here; it is also the 9th fastest out of 18 cards with 16 GB. It runs 14B models well; 27–31B only just fits.

What is the largest model the RX 6950 XT can run?

Gemma 4 26B A4B IT (25.8B) is the largest at 4-bit or better: at Q4_K_S it uses 15.5 GB of 16 GB, and runs at about 74.2 tok/s. At 3-bit, a bigger model fits too: Gemma 2 27B IT (27.2B) at IQ3_M, 15.0 GB.

Can the RX 6950 XT run Qwen3.5 9B?

Yes. Qwen3.5 9B fits at Q4_K_M, using 6.6 GB of 16 GB, and it is the best model that runs well on this card.

Can the RX 6950 XT run 70B models?

No. Llama 3.3 70B Instruct needs 51.3 GB at Q4, more than the 16 GB on this card.

How many tokens per second does the RX 6950 XT give?

About 54.3 tok/s on Qwen3.5 9B at Q4_K_M, with the first token after 7.6 s on a 4K-token prompt.

What power supply does the RX 6950 XT need?

Plan a 700 W PSU (2× the 335 W TDP).

Is 16 GB enough for coding and agents?

Yes for coding: GPT OSS 20B, the best of the 2 coding models that fit, runs at about 81 tok/s. For agents, not yet: no agent-scored model fits in less than 96 GB.

Is renting cheaper than buying an RX 6950 XT?

An RTX 3090 (24 GB) rents from $0.18/hr. Renting costs as much as buying after ≈3,000 hours (≈2.1 years at 4 h a day).

Best coding LLMs 16 GB VRAM 24 GB VRAM GPUs with 16 GB or more RTX 5090 RTX 3090 Ti