Intel · consumer

Arc A310 4GB

Arc A310 4GB has 4 GB of VRAM at 124 GB/s — about 3.72 GiB usable after driver and compositor overhead. 638 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
124 GB/s
64-bit bus
Tensor FP16
—
dense
TDP
75 W
$110 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 507embedding 23vision language 51audio tts 19audio asr 36video 2

What fits at 32K context

largest quantization that fits, per model · 638 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Jan-v1-4BQ2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Jan-nano-128kQ2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Qwen3-4B-Instruct-2507Q2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Qwen3-4B-Thinking-2507Q2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Jan-nanoQ2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Qwen3-4B-abliteratedQ2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
Qwen3-4B-Instruct-2507-hereticQ2_K_L4.0B1.64 GiB1.27 GiB3.72 GiB0.00 GiB23±30%
nomic-embed-codeIQ2_M7.1B2.37 GiB0.49 GiB3.72 GiB0.00 GiB23±30%
gemma-4-E2B-it-abliteratedI1-IQ3_XS5.1B2.85 GiB0.07 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-qat-q4_0-unquantized-hereticI1-IQ3_XS5.1B2.85 GiB0.07 GiB3.71 GiB0.01 GiB23±30%
Huihui-gemma-4-E2B-it-qat-q4_0-unquantized-abliteratedI1-IQ3_XS5.1B2.85 GiB0.07 GiB3.71 GiB0.01 GiB23±30%
Gemma4_E2B_Abliterated_Baked_HF_ReadyI1-IQ3_XS5.1B2.85 GiB0.07 GiB3.71 GiB0.01 GiB23±30%
orpheus-3b-0.1-pretrainedQ2_K_L3.8B1.92 GiB0.98 GiB3.71 GiB0.01 GiB23±30%
Phi-4-mini-reasoningQ3_K_S3.8B1.77 GiB1.13 GiB3.70 GiB0.02 GiB23±30%
Phi-4-mini-instructQ3_K_S3.8B1.77 GiB1.13 GiB3.70 GiB0.02 GiB23±30%
Llama-3.2-3B-Instruct-uncensoredIQ4_XS3.6B1.91 GiB0.98 GiB3.70 GiB0.02 GiB23±30%
EVA-Yi-1.5-9B-32K-V1I1-IQ1_M8.8B2.03 GiB0.84 GiB3.70 GiB0.02 GiB23±30%
Yi-Coder-9B-ChatIQ1_M8.8B2.03 GiB0.84 GiB3.70 GiB0.02 GiB23±30%
OvisOCR2F32853M2.81 GiB0.11 GiB3.70 GiB0.02 GiB23±30%
gemma-4-E2B-itQ4_K_S5.1B2.83 GiB0.07 GiB3.70 GiB0.02 GiB23±30%
umt5-xxlQ3_K_M5.7B2.85 GiB0.00 GiB3.70 GiB0.02 GiB23±30%
whisper-mediumF32764M2.85 GiB0.00 GiB3.69 GiB0.03 GiB23±30%
whisper-medium.enF32764M2.85 GiB0.00 GiB3.69 GiB0.03 GiB23±30%
Voxtral-Mini-3B-2507IQ3_M4.7B1.83 GiB1.05 GiB3.69 GiB0.03 GiB23±30%
SmolLM2-1.7B-Instruct-UncensoredQ5_K_M1.8B1.21 GiB1.69 GiB3.69 GiB0.03 GiB23±30%
Llama-3.2-3B-Instruct-abliteratedI1-IQ4_XS3.6B1.90 GiB0.98 GiB3.69 GiB0.03 GiB23±30%
granite-4.0-h-tiny-baseMoEQ3_K_S6.9B2.85 GiB0.07 GiB3.69 GiB0.03 GiB69±37%
Teuken-7B-instruct-research-v0.4I1-IQ1_M7.5B2.57 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
gemma-3n-E2B-itQ4_K_S5.4B2.77 GiB0.12 GiB3.69 GiB0.03 GiB23±30%
Aura-4BI1-IQ3_XXS4.5B1.75 GiB1.13 GiB3.69 GiB0.03 GiB23±30%
magnum-v2-4bI1-IQ3_XXS4.5B1.75 GiB1.13 GiB3.69 GiB0.03 GiB23±30%
Impish_LLAMA_4BIQ3_XXS4.5B1.75 GiB1.13 GiB3.69 GiB0.03 GiB23±30%
Gemma-3-4b-it-Uncensored-DBL-XI1-Q4_14.7B2.61 GiB0.26 GiB3.69 GiB0.03 GiB23±30%
gemma-4-E2B-it-ultra-uncensored-hereticQ4_K_S5.1B2.82 GiB0.07 GiB3.69 GiB0.03 GiB23±30%
InternVL3_5-8BIQ2_M8.5B2.84 GiB0.00 GiB3.69 GiB0.03 GiB23±30%
FrickFritz-4BI1-Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
qwen3.5-4b-agentic-coder-v4I1-Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Newton-bot-3-VLM-mini-4BQ4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Myth-4BI1-Q4_K_M4.3B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Qwen3.5-4B-UncensoredI1-Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
JOSIE-2-4B-PreviewI1-Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Surogate-3.5-4BI1-Q4_K_M5.3B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Qwopus3.5-4B-Coder-Fable5-v1Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Qwopus3.5-4B-v3Q4_K_M4.7B2.59 GiB0.28 GiB3.69 GiB0.03 GiB23±30%
Luna-7B-A4BMoEI1-IQ1_M6.7B1.61 GiB1.27 GiB3.69 GiB0.03 GiB18±37%
Darwin-4B-ChimeraQ4_K_L4.0B2.51 GiB0.36 GiB3.68 GiB0.04 GiB23±30%
EXAONE-Deep-7.8BQ2_K7.8B2.84 GiB0.00 GiB3.68 GiB0.04 GiB23±30%
EXAONE-3.5-7.8B-InstructQ2_K7.8B2.84 GiB0.00 GiB3.68 GiB0.04 GiB23±30%
alduin-4b-it-baseI1-Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
Qwen3.5-4BQ4_04.7B2.59 GiB0.28 GiB3.68 GiB0.04 GiB23±30%
gemma-3-4b-it-roleplay-tuned-v1I1-Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
Gemma-3-R1984-4BQ5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-ThinkingQ5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
gemma-3-4b-it-roleplay-tuned-v2I1-Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
medgemma-1.5-4b-itQ5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
gemma-3-4b-it-heretic-uncensored-abliterated-ExtremeI1-Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
medgemma-4b-itQ5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
gemma-3-4b-it-abliteratedQ5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
amoral-gemma3-4B-v1Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
Gemma3-4B-CodeCenturionI1-Q5_K_M4.3B2.64 GiB0.22 GiB3.68 GiB0.04 GiB23±30%
●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 Arc A310 4GB run?
638 of 2118 indexed open-weight models fit a Arc A310 4GB at 32,768 context with q4_0 KV cache, the largest being Jan-v1-4B at Q2_K_L. That covers text, vision-language, image, video and speech models.
How much usable memory does a Arc A310 4GB 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 Arc A310 4GB fast for local AI?
Its memory bandwidth is 124 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.