AMD · consumer

Radeon RX 6400

Radeon RX 6400 has 4 GB of VRAM at 128 GB/s — about 3.72 GiB usable after driver and compositor overhead. 585 of 2118 indexed models fit at 32K context with q4_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
53 W
$159 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 460vision language 48audio tts 19audio asr 36video 2embedding 20

What fits at 32K context

largest quantization that fits, per model · 585 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Agents-A1-4B-Heretic-ARA-Refusals8Q4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4B-NSFW-ARA-Heretic-LiteroticaI1-Q4_K_M4.2B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Agents-A1-4BQ4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4B-RpRMax-v1I1-Q4_K_M4.7B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Holo-3.1-4B-uncensored-hereticI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
GRaPE-2-MiniI1-Q4_K_M4.7B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-DPO-4B-2I1-Q4_K_M4.2B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4B-BaseQ4_K_M4.7B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4B-hereticQ4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
NuExtract3Q4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Huihui-Qwen3.5-4B-Claude-4.6-Opus-abliteratedI1-Q4_K_M4.7B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwopus3.5-4B-v3-hereticI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4BQ4_K_M4.7B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Aureth-4B-Qwen3.5I1-Q4_K_M4.5B2.52 GiB0.28 GiB3.72 GiB0.00 GiB28±26.5%
Qwen3.5-4B-Claude-4.6-OS-Auto-Variable-HERETIC-UNCENSORED-THINKINGI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Qwen3.5-4B-SOMPOA-heresy-v2I1-Q4_K_M4.5B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Qwen3.5-4B-SOMPOA-heresyI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Qwen3.5-4B-Safety-ThinkingI1-Q4_K_M4.2B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Huihui-Qwen3.5-4B-abliteratedI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Darkidol-Ballad-4BI1-Q4_K_M4.5B2.52 GiB0.28 GiB3.71 GiB0.01 GiB28±26.5%
Tini-Cybersec-8B-A1BMoEIQ2_M8.5B2.71 GiB0.11 GiB3.71 GiB0.01 GiB72±37%
gemma-3-4b-it-roleplay-tuned-v1I1-Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-ThinkingQ5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
gemma-3-4b-it-roleplay-tuned-v2I1-Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
medgemma-1.5-4b-itQ5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
gemma-3-4b-it-heretic-uncensored-abliterated-ExtremeI1-Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
medgemma-4b-itQ5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
gemma-3-4b-it-abliteratedQ5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
amoral-gemma3-4B-v1Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
Gemma3-4B-CodeCenturionI1-Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
ArrowMint-Gemma3-4B-YUKI-v0.1I1-Q5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
gemma-3-4b-itQ5_K_S4.3B2.57 GiB0.22 GiB3.71 GiB0.01 GiB29±26.5%
gemma-3n-E4B-itUD-IQ2_XXS7.8B2.64 GiB0.14 GiB3.71 GiB0.01 GiB29±26.5%
Nanbeige4.2-3BQ3_K_M4.2B2.02 GiB0.77 GiB3.71 GiB0.01 GiB29±26.5%
Gemma-3-4b-it-Uncensored-DBL-XI1-Q4_K_M4.7B2.52 GiB0.26 GiB3.70 GiB0.02 GiB29±26.5%
Holo-3.1-4BI1-Q3_K_L5.2B2.51 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
AfriqueQwen3.5-4BI1-Q3_K_L5.2B2.51 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
TimeOmni-1-4BI1-Q3_K_L5.2B2.51 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
ToriiGate-0.5Q3_K_L5.2B2.51 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
chandra-ocr-2Q3_K_L5.3B2.51 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
dolphin-2_6-phi-2Q8_02.8B2.75 GiB0.00 GiB3.70 GiB0.02 GiB29±26.5%
Teuken-7B-instruct-research-v0.4I1-IQ1_S7.5B2.48 GiB0.28 GiB3.70 GiB0.02 GiB29±26.5%
dolphin-2.9.3-mistral-7B-32kI1-IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-Instruct-v0.3-ParasiteI1-IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-Instruct-v0.3-JbliteratedI1-IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-Instruct-v0.3IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mathstral-7B-v0.1IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-v0.3IQ1_M7.2B1.64 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Qwen2.5-3B-Instruct-abliteratedI1-Q2_K3.1B2.47 GiB0.32 GiB3.70 GiB0.02 GiB29±26.5%
Darwin-4B-ChimeraQ4_K_M4.0B2.43 GiB0.36 GiB3.70 GiB0.02 GiB29±26.5%
SmolLM2-1.7B-InstructQ5_K_S1.7B1.11 GiB1.69 GiB3.70 GiB0.02 GiB29±26.5%
SciPhi-Self-RAG-Mistral-7B-32kKV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
dolphin-2.2.1-mistral-7bKV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
OpenChat-3.5-7B-Qwen-v2.0KV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
openchat-3.5-0106KV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-Instruct-v0.1KV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Mistral-7B-Instruct-v0.2I1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
ContextualKunoichi_KTO-7BI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
xLAM-7b-rI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 GiB29±26.5%
Ninja-v1-RP-WIPKV unresolvedI1-IQ1_M7.2B1.63 GiB1.13 GiB3.70 GiB0.02 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 RX 6400 run?
585 of 2118 indexed open-weight models fit a Radeon RX 6400 at 32,768 context with q4_0 KV cache, the largest being Agents-A1-4B-Heretic-ARA-Refusals8 at Q4_K_M. That covers text, vision-language, image, video and speech models.
How much usable memory does a Radeon RX 6400 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 RX 6400 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.