Will It Run AI
CalculatorModelsHardwareCompare
Product
  • Calculator
  • Compare
  • Tier List
Browse
  • Models
  • Hardware
  • Docs
About
  • Why It Works
  • What's New
  • Legal Notice
  • Privacy Policy

All estimates are approximations based on mathematical models and public specifications. Actual performance may vary. Do not make purchasing decisions based solely on these estimates.

Data sourced from Hugging Face, Ollama, and official model documentation. Model names and logos are trademarks of their respective owners.

© 2026 Will It Run AI — Fase Consulting Ibiza, S.L. (NIF: B57969656)

Home/Hardware/GPUs/RX 7900 XT 20GB
AMD

AMD

RX 7900 XT 20GB

RX 7000ConsumerRDNA 3PCIe 4ROCm
20GB
VRAM
800GB/s
Bandwidth
52TFLOPS
FP16 Compute
416TOPS
INT8 Inference
$899 MSRP
VRAM20 GBBandwidth800 GB/sCompute52 TFInference416 TOPSValue5.78 TF/$k
RX 7900 XT 20GBCategory AvgMacBook Pro M1 Max 32GB

Specifications

Compute
FP1652 TFLOPS
INT8416 TOPS
ArchitectureRDNA 3
Memory
VRAM20 GB
Bandwidth800 GB/s
General
FamilyRX 7000
SegmentConsumer
InterconnectPCIe 4
Compute PlatformROCM
MSRP$899

Architecture

RDNA 3

RDNA 3 is AMD's chiplet-based GPU architecture, combining a 5nm Graphics Compute Die (GCD) with 6nm Memory Cache Dies (MCDs). It introduces AI accelerators and a new unified compute unit design.

AI Relevance

ROCm support for RDNA 3 is maturing but lags behind NVIDIA's CUDA ecosystem. AI accelerator units provide some inference acceleration, but lack the dedicated Tensor Core equivalent found in NVIDIA GPUs.

Process: TSMC 5nm + 6nmPlatform: ROCMPrecisions: FP32, FP16, BF16, INT8

Recommendations by Workload

Agentic Coding

B

Gemma 3 12B

This model is still usable for agentic-coding, but it is not the most specialized pick. It sits in the middle of the current model mix. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.

Decode 65.6 tok/s · 46K ctx · llama.cpp
14.0 GB / 20.0 GB VRAM

Chat

C

Qwen 3 14B

This model is a direct match for chat. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.

Decode 56.2 tok/s · 13K ctx · llama.cpp
12.5 GB / 20.0 GB VRAM

Coding

C

Qwen 2.5 Coder 14B

This model is a direct match for coding. It sits in the middle of the current model mix. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.

Decode 56.2 tok/s · 23K ctx · llama.cpp
13.6 GB / 20.0 GB VRAM

RAG

C

granite 8b code instruct 4k

This model is a direct match for rag. It sits in the middle of the current model mix. It fits natively with comfortable headroom.

Decode 98.4 tok/s · 62K ctx · llama.cpp
10.3 GB / 20.0 GB VRAM

Reasoning

C

Qwen 3 14B

This model is a direct match for reasoning. It belongs to a current frontier family for local AI. It fits natively with comfortable headroom. Known channels: huggingface, ollama, lm-studio.

Decode 56.2 tok/s · 23K ctx · llama.cpp
13.6 GB / 20.0 GB VRAM

Full Model Compatibility

UnslothQwen3.5 9B
C53
9B9.8 GB87 tok/s33K ctx
dense
HauhauCSHQwen3.5 9B Uncensored HauhauCS Aggressive
C53
9B9.8 GB87 tok/s33K ctx
dense
Lmstudio-communityLQwen3.5 9B
C52
9B9.8 GB87 tok/s33K ctx
dense
BartowskiBMeta Llama 3.1 8B Instruct
C52
8B9.0 GB98 tok/s35K ctx
dense
XtunerXllava llama 3 8b v1 1
C52
8B9.0 GB98 tok/s35K ctx
dense
UnslothDeepSeek R1 0528 Qwen3 8B
C52
8B9.0 GB98 tok/s35K ctx
dense
MaziyarPanahiMMeta Llama 3 8B Instruct
C52
8B9.0 GB98 tok/s35K ctx
dense
TheBlokeTLlama 2 7B Chat
C51
7B8.3 GB112 tok/s39K ctx
dense
MistralCodestral 2 25.08
C51
22B19.8 GB36 tok/s16K ctx
dense
TheBlokeTMistral 7B Instruct v0.2
C51
7B8.3 GB112 tok/s39K ctx
dense
MaziyarPanahiMMistral 7B Instruct v0.3
C51
7B8.3 GB112 tok/s39K ctx
dense
UnslothQwen3.5 4B
C49
4B6.1 GB197 tok/s52K ctx
dense
Lmstudio-communityLgemma 3 4b it
C49
4B6.1 GB197 tok/s52K ctx
dense
BartowskiBLlama 3.2 3B Instruct
C49
3B5.9 GB227 tok/s55K ctx
dense
QwenQwen2.5 3B Instruct
C48
3B5.5 GB262 tok/s58K ctx
dense
BartowskiBgemma 2 2b it
C48
2B5.3 GB307 tok/s60K ctx
dense
Googlegemma 2b
C47
2B4.9 GB393 tok/s65K ctx
dense
TheDrummerTGemmasutra Mini 2B v1
C47
2B4.9 GB393 tok/s65K ctx
dense
QwenQwen2.5 1.5B Instruct
C47
1.5B4.6 GB480 tok/s69K ctx
dense
Hugging-quantsHLlama 3.2 1B Instruct Q8 0
C47
1B4.5 GB504 tok/s71K ctx
dense
TheBlokeTTinyLlama 1.1B Chat v1.0
C47
1.1B4.4 GB480 tok/s73K ctx
dense
Ggml-orgGSmolVLM 500M Instruct
C47
0.5B4.1 GB504 tok/s78K ctx
dense
Ggml-orgGembeddinggemma 300M
C46
0.3B3.9 GB504 tok/s81K ctx
dense
AlibabaQwen3-VL 30B A3B Instruct
C43
30B22.0 GB64 tok/s15K ctx
moe
AlibabaQwen3-Coder 30B A3B Instruct
C43
30.5B22.3 GB61 tok/s14K ctx
moe
MistralDevstral Small 2 24B Instruct
C41
24B21.3 GB31 tok/s15K ctx
dense
MistralDevstral Small 1.1
C41
24B21.3 GB31 tok/s15K ctx
dense
UnslothQwen3.5 27B
D34
27B23.6 GB25 tok/s14K ctx
dense
Unslothgemma 3 27b it
D34
27B23.6 GB25 tok/s14K ctx
dense
DeepSeekDeepSeek R1 671B
F0
671B418.0 GB3 tok/s4K ctx
moe
MistralDevstral 2 123B Instruct
F0
123B97.1 GB6 tok/s4K ctx
dense
Z.aiGLM-5
F0
744B463.0 GB3 tok/s4K ctx
moe
UnslothQwen3.5 35B A3B
F0
35B29.7 GB23 tok/s11K ctx
dense
Moonshot AIKimi K2.5
F0
1000B617.9 GB2 tok/s4K ctx
moe
MistralMistral Large 3
F0
675B421.1 GB3 tok/s4K ctx
+1moe
MistralMistral Small 4 119B
F0
119B76.5 GB19 tok/s4K ctx
moe
AlibabaQwen3-Coder 480B A35B Instruct
F0
480B301.2 GB5 tok/s4K ctx
moe
AlibabaQwen3-Coder-Next
F0
80B52.5 GB30 tok/s6K ctx
moe
UnslothQwen3.5 122B A10B
F0
122B81.7 GB8 tok/s4K ctx
dense
DeepSeekDeepSeek V3 671B
F0
671B418.0 GB3 tok/s4K ctx
moe
MistralMixtral 8x22B
F0
141B95.0 GB11 tok/s4K ctx
moe
AlibabaQwen 2.5 72B
F0
72B58.1 GB11 tok/s6K ctx
dense
AlibabaQwen 3 235B A22B
F0
235B149.7 GB9 tok/s4K ctx
moe
UnslothQwen3.5 397B A17B
F0
397B307.1 GB2 tok/s4K ctx
dense
MetaLlama 3.3 70B
F0
70B56.5 GB11 tok/s6K ctx
dense
MetaLlama 4 Maverick 17B 128E
F0
400B249.6 GB6 tok/s4K ctx
moe
CohereCommand A 111B
F0
111B88.0 GB7 tok/s4K ctx
dense
AlibabaQwen 2.5 Coder 32B
F0
32B27.4 GB25 tok/s12K ctx
dense
AlibabaQwen 2.5 VL 72B
F0
72B58.1 GB11 tok/s6K ctx
dense
Lmstudio-communityLQwen3.5 35B A3B
F0
35B29.7 GB23 tok/s11K ctx
dense

Just out of reach

Models you could run with an upgrade

High-quality models that need a bit more memory

DeepSeekDeepSeek R1 671B
671BTier 5Needs ~423.8 GB
MistralDevstral 2 123B Instruct
123BTier 5Needs ~116.4 GB
Runs on Mac Studio M3 Ultra 256GB
Z.aiGLM-5
744BTier 5Needs ~469.2 GB
UnslothQwen3.5 27B
27BTier 5Needs ~27.8 GB
Runs on RTX 5090 32GB (~$1,999)
UnslothQwen3.5 35B A3B
35BTier 5Needs ~35.2 GB
Runs on Mac mini M4 64GB (~$1,099)

Upgrade paths

Upgrade from RX 7900 XT 20GB

See what you unlock with more powerful hardware

Upgrade options

Upgrade options

AppleMacBook Pro M1 Max 32GBNext step up
32 GB Unified (+12)
A
Unlocks Devstral Small 2 24B Instruct, Devstral Small 1.1, Mistral Small 3.2 24B Instruct 2506+9 more

~$2,499 MSRP

AMDAMD Instinct MI100 32GBAMD upgrade
32 GB VRAM (+12)1228 GB/s (+428)
A
Unlocks Qwen3.5 27B, Qwen3.5 35B A3B, Devstral Small 2 24B Instruct+31 more · +50% faster avg

 

AppleMac mini M4 64GBBest value
64 GB Unified (+44)
B
Unlocks Qwen3.5 27B, Qwen3.5 35B A3B, Devstral Small 2 24B Instruct+33 more

~$1,099 MSRP

AMDAMD Instinct MI350X 288GBBiggest leap
288 GB VRAM (+268)8000 GB/s (+7200)
A
Unlocks Devstral 2 123B Instruct, Qwen3.5 27B, Qwen3.5 35B A3B+63 more · +898% faster avg

~$8,000 MSRP

Compare this GPU