AMD · workstation

Radeon Pro W6400

Radeon Pro W6400 has 4 GB of VRAM at 128 GB/s — about 3.72 GiB usable after driver and compositor overhead. 772 of 2118 indexed models fit at 8K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
4 GB
GDDR6
Bandwidth
128 GB/s
64-bit bus
Tensor FP16
—
dense
TDP
50 W
$229 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 635vision language 54embedding 25audio asr 37audio tts 19video 2

What fits at 8K context

largest quantization that fits, per model · 772 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Darwin-4B-ChimeraQ4_K_M4.0B2.43 GiB0.38 GiB3.72 GiB0.00 GiB28±26.5%
OpenCaption-4B-VL-SFT-v1.0Q4_04.4B2.21 GiB0.60 GiB3.72 GiB0.00 GiB28±26.5%
gemma-4-E2B-itIQ4_XS5.1B2.78 GiB0.04 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3-Reranker-4BQ4_04.0B2.21 GiB0.60 GiB3.72 GiB0.00 GiB28±26.5%
Holo-3.1-4BIQ4_XS5.2B2.67 GiB0.13 GiB3.71 GiB0.01 GiB29±26.5%
ToriiGate-0.5IQ4_XS5.2B2.67 GiB0.13 GiB3.71 GiB0.01 GiB29±26.5%
Yi-6B-ChatI1-Q3_K_S6.1B2.52 GiB0.27 GiB3.71 GiB0.01 GiB29±26.5%
Yi-1.5-6B-ChatQ3_K_S6.1B2.52 GiB0.27 GiB3.71 GiB0.01 GiB29±26.5%
gemma-4-E2B-it-qat-q4_0-unquantized-hereticI1-Q2_K5.1B2.78 GiB0.04 GiB3.71 GiB0.01 GiB28±26.5%
Huihui-gemma-4-E2B-it-qat-q4_0-unquantized-abliteratedI1-Q2_K5.1B2.78 GiB0.04 GiB3.71 GiB0.01 GiB28±26.5%
Gemma4_E2B_Abliterated_Baked_HF_ReadyI1-Q2_K5.1B2.78 GiB0.04 GiB3.71 GiB0.01 GiB28±26.5%
MiMo-VL-7B-RLI1-IQ2_XXS8.3B2.18 GiB0.60 GiB3.71 GiB0.01 GiB29±26.5%
Kuwutu-7B-CYOA-v2I1-IQ2_XXS7.6B2.18 GiB0.60 GiB3.71 GiB0.01 GiB29±26.5%
Foundation-Sec-8B-InstructI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Foundation-Sec-8B-Instruct-hereticI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Meta-Llama-3-8BIQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
llama3.1-heretic-unsensoredI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Anubis-Mini-8B-v1I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
dolphin-2.9-llama3-8bIQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Meta-Llama-3-8BIQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
L3.1-Dark-Reasoning-LewdPlay-evo-Hermes-R1-Uncensored-8B-hereticI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
L3.1-Dark-Reasoning-LewdPlay-evo-Hermes-R1-Uncensored-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Gluon-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-ReasoningI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama3.3-8B-Instruct-Thinking-Claude-4.5-Opus-High-ReasoningI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Hypnos-i1-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
grok-oss-Apollyon-8B-hereticI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.3-8B-Instruct-128K-JbliteratedI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
PsyCoPref-Llama3-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama3.1-GptDeluxe-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
SelfCite-8B-CC-SFTI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
ai-girlfriend-v2I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.3-8B-Instruct-128K_AbliteratedI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Aisha-Uncensored-v2I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
grok-oss-Apollyon-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
HARC-Llama-3.1-8B-InstructI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Lumen-1.2.5I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.3-8B-Instruct-128KI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Stoicism1_Llama3.1-8b-instructI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
grok-oss-Thanatos-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
OpenElla-NovelWriter-Requiem-AscendedI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
llama-joycaption-beta-one-hf-llavaI1-IQ2_XXS8.5B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.1-Swallow-8B-Instruct-v0.5IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
DarkIdol-Llama-3.1-8B-Instruct-1.3-UncensoredI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
calme-2.3-legalkit-8bI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama3.1-DilemmaDeluxe-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.1-8B-Stheno-v3.4I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
DarkIdol-Llama-3.1-8B-Instruct-1.0-UncensoredI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
SzilviaB-Daredevil-LongWriter-8B_abliteratedI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.1-SuperNova-LiteI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Humanish-Roleplay-Llama-3.1-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.1_OpenScholar-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Meta-Llama-3.1-8B-ClaudeI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.1-Storm-8BIQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Meta-Llama-3.1-8B-Instruct-abliteratedI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-3.3-8B-InstructI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Infinity-Instruct-7M-Gen-Llama3_1-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
markdownizerI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Llama-SEA-LION-v2-8BI1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
Meta-Llama-3-8B-Instruct-abliterated-v3I1-IQ2_XXS8.0B2.23 GiB0.53 GiB3.71 GiB0.01 GiB29±26.5%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a Radeon Pro W6400 run?
772 of 2118 indexed open-weight models fit a Radeon Pro W6400 at 8,192 context with q8_0 KV cache, the largest being Darwin-4B-Chimera at Q4_K_M. That covers text, vision-language, image, video and speech models.
How much usable memory does a Radeon Pro W6400 actually have?
Its nameplate is 4 GB, but about 3.72 GiB is available to a model once driver and compositor overhead is accounted for.
Is a Radeon Pro W6400 fast for local AI?
Its memory bandwidth is 128 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.