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 6900 XT 16GB
AMD

AMD

RX 6900 XT 16GB

RX 6000ConsumerRDNA 2PCIe 4ROCm
16GB
VRAM
512GB/s
Bandwidth
46TFLOPS
FP16 Compute
368TOPS
INT8 Inference
VRAM16 GBBandwidth512 GB/sCompute46 TFInference368 TOPS
RX 6900 XT 16GBCategory AvgRTX 4000 Ada 20GB

Specifications

Compute
FP1646 TFLOPS
INT8368 TOPS
ArchitectureRDNA 2
Memory
VRAM16 GB
Bandwidth512 GB/s
General
FamilyRX 6000
SegmentConsumer
InterconnectPCIe 4
Compute PlatformROCM

Architecture

RDNA 2

RDNA 2 is AMD's second-generation RDNA architecture, built on TSMC 7nm. It introduced hardware ray tracing and Infinity Cache for improved bandwidth efficiency. Powers the RX 6000 series and is also used in gaming consoles.

AI Relevance

Limited official ROCm support for consumer RDNA 2 cards — most AI runtimes require workarounds. Can run smaller models via llama.cpp with Vulkan or HIP backends, but performance is well behind NVIDIA equivalents.

Process: TSMC 7nmPlatform: ROCMPrecisions: FP32, FP16, INT8

Recommendations by Workload

Agentic Coding

C

Yi Coder 9B

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 53.2 tok/s · 47K ctx · llama.cpp
10.8 GB / 16.0 GB VRAM

Chat

C

Qwen 3 8B

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 59.8 tok/s · 16K ctx · llama.cpp
8.2 GB / 16.0 GB VRAM

Coding

C

Yi Coder 9B

This model is still usable for 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 53.2 tok/s · 27K ctx · llama.cpp
9.4 GB / 16.0 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 59.8 tok/s · 52K ctx · llama.cpp
9.9 GB / 16.0 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 should run, but memory headroom will be limited. Known channels: huggingface, ollama, lm-studio.

Decode 34.2 tok/s · 19K ctx · llama.cpp
13.2 GB / 16.0 GB VRAM

Full Model Compatibility

UnslothQwen3.5 9B
C53
9B9.4 GB53 tok/s27K ctx
dense
HauhauCSHQwen3.5 9B Uncensored HauhauCS Aggressive
C53
9B9.4 GB53 tok/s27K ctx
dense
Lmstudio-communityLQwen3.5 9B
C53
9B9.4 GB53 tok/s27K ctx
dense
BartowskiBMeta Llama 3.1 8B Instruct
C52
8B8.6 GB60 tok/s30K ctx
dense
XtunerXllava llama 3 8b v1 1
C52
8B8.6 GB60 tok/s30K ctx
dense
UnslothDeepSeek R1 0528 Qwen3 8B
C52
8B8.6 GB60 tok/s30K ctx
dense
MaziyarPanahiMMeta Llama 3 8B Instruct
C52
8B8.6 GB60 tok/s30K ctx
dense
TheBlokeTLlama 2 7B Chat
C52
7B7.9 GB68 tok/s33K ctx
dense
TheBlokeTMistral 7B Instruct v0.2
C52
7B7.9 GB68 tok/s33K ctx
dense
MaziyarPanahiMMistral 7B Instruct v0.3
C52
7B7.9 GB68 tok/s33K ctx
dense
UnslothQwen3.5 4B
C50
4B5.7 GB120 tok/s45K ctx
dense
Lmstudio-communityLgemma 3 4b it
C50
4B5.7 GB120 tok/s45K ctx
dense
BartowskiBLlama 3.2 3B Instruct
C50
3B5.5 GB138 tok/s47K ctx
dense
QwenQwen2.5 3B Instruct
C49
3B5.1 GB160 tok/s50K ctx
dense
BartowskiBgemma 2 2b it
C49
2B4.9 GB187 tok/s52K ctx
dense
Googlegemma 2b
C48
2B4.5 GB239 tok/s57K ctx
dense
TheDrummerTGemmasutra Mini 2B v1
C48
2B4.5 GB239 tok/s57K ctx
dense
QwenQwen2.5 1.5B Instruct
C48
1.5B4.2 GB292 tok/s61K ctx
dense
Hugging-quantsHLlama 3.2 1B Instruct Q8 0
C48
1B4.1 GB306 tok/s62K ctx
dense
TheBlokeTTinyLlama 1.1B Chat v1.0
C47
1.1B4.0 GB292 tok/s64K ctx
dense
Ggml-orgGSmolVLM 500M Instruct
C47
0.5B3.7 GB306 tok/s69K ctx
dense
Ggml-orgGembeddinggemma 300M
C47
0.3B3.5 GB306 tok/s72K ctx
dense
DeepSeekDeepSeek R1 671B
F0
671B417.6 GB2 tok/s4K ctx
moe
MistralDevstral 2 123B Instruct
F0
123B96.7 GB4 tok/s4K ctx
dense
Z.aiGLM-5
F0
744B462.6 GB2 tok/s4K ctx
moe
UnslothQwen3.5 27B
F0
27B23.2 GB18 tok/s11K ctx
dense
UnslothQwen3.5 35B A3B
F0
35B29.3 GB14 tok/s9K ctx
dense
Moonshot AIKimi K2.5
F0
1000B617.5 GB2 tok/s4K ctx
moe
MistralMistral Large 3
F0
675B420.7 GB2 tok/s4K ctx
+1moe
MistralMistral Small 4 119B
F0
119B76.1 GB12 tok/s4K ctx
moe
AlibabaQwen3-Coder 30B A3B Instruct
F0
30.5B21.9 GB41 tok/s12K ctx
moe
AlibabaQwen3-Coder 480B A35B Instruct
F0
480B300.8 GB3 tok/s4K ctx
moe
AlibabaQwen3-Coder-Next
F0
80B52.1 GB18 tok/s5K ctx
moe
UnslothQwen3.5 122B A10B
F0
122B81.3 GB5 tok/s4K ctx
dense
DeepSeekDeepSeek V3 671B
F0
671B417.6 GB2 tok/s4K ctx
moe
MistralMixtral 8x22B
F0
141B94.6 GB7 tok/s4K ctx
moe
AlibabaQwen 2.5 72B
F0
72B57.7 GB7 tok/s4K ctx
dense
AlibabaQwen 3 235B A22B
F0
235B149.3 GB5 tok/s4K ctx
moe
AlibabaQwen3-VL 30B A3B Instruct
F0
30B21.6 GB42 tok/s12K ctx
moe
UnslothQwen3.5 397B A17B
F0
397B306.7 GB2 tok/s4K ctx
dense
MistralDevstral Small 2 24B Instruct
F0
24B20.9 GB20 tok/s12K ctx
dense
MetaLlama 3.3 70B
F0
70B56.1 GB7 tok/s5K ctx
dense
MetaLlama 4 Maverick 17B 128E
F0
400B249.2 GB4 tok/s4K ctx
moe
CohereCommand A 111B
F0
111B87.6 GB4 tok/s4K ctx
dense
AlibabaQwen 2.5 Coder 32B
F0
32B27.0 GB15 tok/s9K ctx
dense
AlibabaQwen 2.5 VL 72B
F0
72B57.7 GB7 tok/s4K ctx
dense
Unslothgemma 3 27b it
F0
27B23.2 GB18 tok/s11K ctx
dense
Lmstudio-communityLQwen3.5 35B A3B
F0
35B29.3 GB14 tok/s9K ctx
dense
MistralCodestral 2 25.08
F0
22B19.4 GB22 tok/s13K ctx
dense
MistralDevstral Small 1.1
F0
24B20.9 GB20 tok/s12K 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 ~423.4 GB
MistralDevstral 2 123B Instruct
123BTier 5Needs ~116.0 GB
Runs on Mac Studio M3 Ultra 256GB
Z.aiGLM-5
744BTier 5Needs ~468.8 GB
UnslothQwen3.5 27B
27BTier 5Needs ~27.4 GB
Runs on RTX 5090 32GB (~$1,999)
UnslothQwen3.5 35B A3B
35BTier 5Needs ~34.8 GB
Runs on Mac mini M4 64GB (~$1,099)

Upgrade paths

Upgrade from RX 6900 XT 16GB

See what you unlock with more powerful hardware

Upgrade options

Upgrade options

NVIDIARTX 4000 Ada 20GBNext step up
20 GB VRAM (+4)
A
Unlocks Qwen3-Coder 30B A3B Instruct, Qwen3-VL 30B A3B Instruct, Codestral 2 25.08+15 more

 

AMDRX 7900 XTX 24GBAMD upgrade
24 GB VRAM (+8)960 GB/s (+448)
B
Unlocks Qwen3.5 27B, Qwen3-Coder 30B A3B Instruct, Qwen3-VL 30B A3B Instruct+32 more · +113% faster avg

~$999 MSRP

AppleMac mini M4 64GBBest value
64 GB Unified (+48)
B
Unlocks Qwen3.5 27B, Qwen3.5 35B A3B, Qwen3-Coder 30B A3B Instruct+51 more

~$1,099 MSRP

AMDAMD Instinct MI350X 288GBBiggest leap
288 GB VRAM (+272)8000 GB/s (+7488)
A
Unlocks Devstral 2 123B Instruct, Qwen3.5 27B, Qwen3.5 35B A3B+81 more · +1453% faster avg

~$8,000 MSRP

Compare this GPU