Best AI Models for AMD Radeon RX 9070 (16 GB)

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

16 GB GDDR6640 GB/s145 TFLOPS FP16220 WRDNA 4 2025

What LLMs can the RX 9070 run?

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 118 t/s	118 tok/s	Q4_K_M · 3.4 GB
3	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 86.6 t/s	86.6 tok/s	Q4_K_M · 13.1 GB
4	Phi 4Microsoft Q4_K_M · 12.5 GB · 40.3 t/s	40.3 tok/s	Q4_K_M · 12.5 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 46.7 t/s	46.7 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 · 86.6 t/s	86.6 tok/s		Q4_K_M · 13.1 GB
2	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 72.5 t/s	72.5 tok/s		Q4_K_M · 7.9 GB
3	Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	overall	Q4_K_M · 6.6 GB
4	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 46.7 t/s	46.7 tok/s		Q4_K_M · 9.0 GB
5	MiMo V2.6 Distill Qwen 9BXiaomi Q4_K_M · 6.5 GB · 61.8 t/s	61.8 tok/s		Q4_K_M · 6.5 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	Q4_K_M · 6.6 GB
2	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 86.6 t/s	86.6 tok/s	Q4_K_M · 13.1 GB
3	Qwen3 8BAlibaba Q4_K_M · 7.6 GB · 69.6 t/s	69.6 tok/s	Q4_K_M · 7.6 GB
4	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 72.5 t/s	72.5 tok/s	Q4_K_M · 7.9 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 46.7 t/s	46.7 tok/s	Q4_K_M · 9.0 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 118 t/s	118 tok/s	Q4_K_M · 3.4 GB
3	Mistral Nemo Instruct 2407Mistral AI Q4_K_M · 11.0 GB · 47.6 t/s	47.6 tok/s	Q4_K_M · 11.0 GB
4	Llama 3.1 8B InstructMeta Q4_K_M · 7.9 GB · 72.5 t/s	72.5 tok/s	Q4_K_M · 7.9 GB
5	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 46.7 t/s	46.7 tok/s	Q4_K_M · 9.0 GB

#	Model	Speed ↓	Quant · VRAM
1	Qwen3.5 4BAlibabaHow to run → Q4_K_M · 3.4 GB · 118 t/s	118 tok/s	Q4_K_M · 3.4 GB
2	GPT OSS 20BOpenAI Q4_K_M · 13.1 GB · 86.6 t/s	86.6 tok/s	Q4_K_M · 13.1 GB
3	Qwen3.5 9BAlibaba Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	Q4_K_M · 6.6 GB
4	Gemma 4 12B ITGoogle Q4_K_M · 9.0 GB · 46.7 t/s	46.7 tok/s	Q4_K_M · 9.0 GB
5	Phi 4Microsoft Q4_K_M · 12.5 GB · 40.3 t/s	40.3 tok/s	Q4_K_M · 12.5 GB

#	Model	Speed	Quant · VRAM
1	Qwen3.5 9BAlibabaHow to run → Q4_K_M · 6.6 GB · 60.4 t/s	60.4 tok/s	Q4_K_M · 6.6 GB
2	Qwen3.5 4BAlibaba Q4_K_M · 3.4 GB · 118 t/s	118 tok/s	Q4_K_M · 3.4 GB
3	Gemma 3 12B ITGoogle Q4_K_M · 10.9 GB · 47.8 t/s	47.8 tok/s	Q4_K_M · 10.9 GB
4	Gemma 3 4B ITGoogle Q4_K_M · 5.0 GB · 135 t/s	135 tok/s	Q4_K_M · 5.0 GB
5	MiMo V2.6 Distill Qwen 9BXiaomi Q4_K_M · 6.5 GB · 61.8 t/s	61.8 tok/s	Q4_K_M · 6.5 GB

See all 307 models on the RX 9070 →Best quant that fits · speeds estimated on this card

About the RX 9070

The RX 9070 is a solid mid-range card in the 16 GB class: it runs 14B models well. It is among the most efficient per watt in its tier, and fair value at ≈$800. RDNA 4 generation, released 2025.

Speed in 16 GB#6640 GB/s

Per watt#20.27 tok/s per W on Qwen3.5 9B (9B params)

Per dollar#47.6 tok/s per $100 on Qwen3.5 9B (9B params)

GenerationRDNA 4released 2025

RX 9070 strengths and weaknesses

Strengths

Runs 35B MoE models: Qwen3.6 35B A3B at 74.7 tok/s in 15.7 GB
Reads long prompts fast: A 4K-token prompt on Qwen3.5 9B starts answering in 2.5 s

Weaknesses

70B models don't fit: Llama 3.3 70B Instruct needs 51.3 GB at Q4 vs 16 GB here
ROCm, not CUDA: No FP8 weights; MXFP4 only through llama.cpp, not vLLM
Priced well above MSRP: ≈$800 new vs its $549 MSRP (+46%)

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.

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
RX 7900 XTX24 GB	about $1,130	90.6	8.0	25.5
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
RTX 508016 GB	about $1,690	98.1	5.8	27.3
RX 9060 XT 16GB16 GB	about $590	30.2	5.1	18.9
RTX 509032 GB	about $5,376	183	3.4	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
RTX 507012 GB	about $736	68.7	9.3	27.5
RX 907016 GB	about $650	60.4	9.3	27.5
RX 7900 XTX24 GB	about $1,000	90.6	9.1	25.5
RTX 5070 Ti16 GB	about $1,100	91.6	8.3	30.5
RX 9070 XT16 GB	about $790	60.4	7.6	19.9
RTX 4070 Ti SUPER16 GB	about $900	68.7	7.6	24.1
RX 9060 XT 16GB16 GB	about $480	30.2	6.3	18.9

Cost per token: the RX 9070 against 12–32 GB cards

On Qwen3.5 9B (9B params, Q4_K_M, 4K context). Speed, cost per token and power per token depend on the model.

12 GB16 GB24 GB32 GBthis card

Cost per 1M output tokenson Qwen3.5 9B · card over 3 years + power

$0.99

Card, new price: $0.84
Power: $0.15

3-year cost of ownership$945

Or rent an RTX 3090 in the cloud$0.19/hr

Renting costs as much as buying after ≈4,200 hours (≈2.9 years at 4 h a day).

Include card price

Power$/kWhUseh/dayKeepyears

Each dot is a card; hover or tap it for its numbers, click to open it. Cards without a price for the chosen basis are left out. How we calculate cost per token

Similar GPUs

Cards people weigh against the RX 9070, each with the one reason it is worth a look. Speeds on Qwen3.5 9B.

14% faster, same 16 GBRTX 4070 Ti SUPERNVIDIA · 16 GB · Ada Lovelace

▲ 14% speedsame 16 GB$799rent $0.15/hr

52% faster for $370 moreRTX 5070 TiNVIDIA · 16 GB · Blackwell

▲ 52% speed▲ $370 moresame 16 GB$1,170rent $0.16/hr

Slower but $210 cheaperRX 9060 XT 16GBAMD · 16 GB · RDNA 4

▼ 50% speed▼ $210 lesssame 16 GB$590

8 GB more, 50% faster for $330 moreRX 7900 XTXAMD · 24 GB · RDNA 3

▲ 50% speed▲ $330 more▲ 8 GB$1,130

39% slower, 12 GBRTX 3060 12GBNVIDIA · 12 GB · Ampere

▼ 39% speed▼ 4 GB$329rent $0.06/hr

4 GB more and 25% fasterRX 7900 XTAMD · 20 GB · RDNA 3

▲ 25% speed▲ 4 GB$899

RX 9070 tokens per second by model size

Output speed depends on memory bandwidth (640 GB/s). Grey rows need more than 16 GB.

9BQwen3.5 9B · Q4_K_M · 6.6 GB
60.4tok/s
14BQwen3 14B · Q4_K_M · 11.9 GB
40.4tok/s
27–31BQwen3.8 27B · Q3_K_L · 15.4 GB tight · Q3_K_L
25.6tok/s
35B MoEQwen3.6 35B A3B · IQ3_S · 15.7 GB tight · IQ3_S
74.7tok/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 1057 models on the RX 9070

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 9070 — ranked by Score, recency and Fit
Model	Quant	VRAM	Speed	Context	Fit
Llama2 0B Unit Test770940Chat Q4_K_M·1280.0 t/s	Q4_K_M	0.3 GB2%	1280.0 t/s	1K	S100, Runs great
Qwen3 Coder 30B A3B Instruct30.5BChatCodeFunctions Q3_K_M·71.6 t/s	Q3_K_M	16.0 GB100%	71.6 t/s	262K	C44, Usable
Hyenadna Medium 450k Seqlen HF28MChat BF16·6400.0 t/s	BF16	0.6 GB4%	6400.0 t/s	450K	S100, Runs great
Qwen3 30B A3B Thinking 250730.5BChatFunctions Q3_K_M·71.6 t/s	Q3_K_M	16.0 GB100%	71.6 t/s	262K	C44, Usable
Gpt2 Small Dutch129MChat Q4_K_M·4266.7 t/s	Q4_K_M	0.1 GB1%	4266.7 t/s	1K	S100, Runs great
SmolLM2 360M Instruct362MChat Q4_K_M·640.0 t/s	Q4_K_M	0.8 GB5%	640.0 t/s	8K	S100, Runs great
EXAONE 3.5 2.4B Instruct2.4BChat Q4_K_M·241.5 t/s	Q4_K_M	3.4 GB21%	241.5 t/s	33K	S100, Runs great
Natural Sql 7B6.9BChat Q4_K_M·70.5 t/s	Q4_K_M	6.5 GB40%	70.5 t/s	4K	S86, Runs great
Qwen3.6 27B Uncensored HauhauCS Aggressive Safetensor Benchmark27.8BChat Q3_K_M·26.9 t/s	Q3_K_M	14.7 GB92%	26.9 t/s	262K	B50, Decent
Pythia 410M506MChat Q4_K_M·1163.6 t/s	Q4_K_M	0.3 GB2%	1163.6 t/s	2K	S100, Runs great
Gpt2 Xl1.6BChat Q4_K_M·362.3 t/s	Q4_K_M	1.1 GB7%	362.3 t/s	1K	S100, Runs great
GOT OCR2 0716MVision Q4_K_M·412.9 t/s	Q4_K_M	2.3 GB15%	412.9 t/s	33K	S100, Runs great
Pythia 1.4B1.5BChat Q4_K_M·384.0 t/s	Q4_K_M	1 GB6%	384.0 t/s	2K	S100, Runs great
GLM 4.1V 9B Thinking10.3BVisionReasoning Q4_K_M·58.5 t/s	Q4_K_M	7.2 GB45%	58.5 t/s	66K	A83, Runs well
Qwen2.5 14B Instruct14.8BChatFunctions Q4_K_M·40.2 t/s	Q4_K_M	12.4 GB77%	40.2 t/s	33K	A75, Runs well

RX 9070 specs, power and PSU

Specs

Architecture: RDNA 4
VRAM: 16 GB GDDR6
Memory bandwidth: 640.0 GB/s
FP16 (dense): 145.00 TFLOPS
Stream processors: 3,584
TDP: 220 W
Released: 2025-03-06
MSRP: $549

RX 9070 FAQ

Is the RX 9070 still worth it in 2026?

At ≈$800 you get 7.6 tok/s per $100 on Qwen3.5 9B, which ranks 4th of the 8 cards compared here; it is also the 6th fastest out of 17 cards with 16 GB. It runs 14B models well; 27–31B only just fits.

What is the largest model the RX 9070 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 78.6 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 9070 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 9070 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 9070 give?

About 60.4 tok/s on Qwen3.5 9B at Q4_K_M, with the first token after 2.5 s on a 4K-token prompt.

What power supply does the RX 9070 need?

Plan a 450 W PSU (2× the 220 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 87 tok/s. For agents, not yet: no agent-scored model fits in less than 96 GB.

Is renting cheaper than buying an RX 9070?

An RTX 3090 (24 GB) rents from $0.19/hr. Renting costs as much as buying after ≈4,200 hours (≈2.9 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