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. 513 of 2118 indexed models fit at 16K context with f16 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 393vision language 45audio asr 35video 2embedding 20audio tts 18

What fits at 16K context

largest quantization that fits, per model · 513 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Qwen3-Zero-Coder-Reasoning-V2-0.8BI1-IQ2_M816M0.32 GiB2.63 GiB3.72 GiB0.00 GiB22±30%
Supertron2-Reranker-2BI1-Q5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Uni-MuMER-Qwen3-VL-2BI1-Q5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Qwen3-VL-2B-ThinkingQ5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Qwen3-VL-Reranker-2BI1-Q5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Qwen3-VL-2B-InstructQ5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Qwen3-VL-Embedding-2BQ5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
OpenCaption-2B-VL-SFT-v1.0I1-Q5_K_M2.1B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Atomight-V2.5-1.7BI1-Q5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
OpenClaude-1.7B-MergedI1-Q5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
gaon-1.7b-v2-translateI1-Q5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
gaon-1.7b-v2-instructI1-Q5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Lightning-1.7BQ5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
DorsetHeatwaveLLM2I1-Q5_K_M1.7B1.17 GiB1.75 GiB3.72 GiB0.00 GiB23±30%
Felldude-Uncensored-Ministral3-3B-bf16I1-Q2_K_S3.8B1.28 GiB1.63 GiB3.71 GiB0.01 GiB23±30%
Amaretto-3BI1-Q2_K_S4.3B1.28 GiB1.63 GiB3.71 GiB0.01 GiB23±30%
FrickFritz-4BI1-IQ4_XS4.7B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
qwen3.5-4b-agentic-coder-v4I1-IQ4_XS4.7B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
Myth-4BI1-IQ4_XS4.3B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
Qwen3.5-4B-UncensoredI1-IQ4_XS4.7B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
JOSIE-2-4B-PreviewI1-IQ4_XS4.7B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
Surogate-3.5-4BI1-IQ4_XS5.3B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-ultra-uncensored-hereticI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Barcenas-E2BI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-Uncensored-MAXI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Firefly-v4I1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-abliteratedI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-uncensoredQ2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-heretic-araQ2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
rp-model_E2B_v5.1I1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
BartaLens-E2BI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2BQ2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Qwen3.5-4BQ3_K_M4.7B2.40 GiB0.50 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-itIQ4_XS5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
granite-3.1-2b-instructQ5_K_S2.5B1.66 GiB1.25 GiB3.71 GiB0.01 GiB23±30%
gemma-4-E2B-it-qat-q4_0-unquantized-hereticI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Huihui-gemma-4-E2B-it-qat-q4_0-unquantized-abliteratedI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Gemma4_E2B_Abliterated_Baked_HF_ReadyI1-Q2_K5.1B2.78 GiB0.14 GiB3.71 GiB0.01 GiB23±30%
Llama-3.2-3B-Instruct-abliteratedI1-IQ2_XS3.6B1.15 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Dolphin3.0-Llama3.2-3BIQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Llama-Doctor-3.2-3B-InstructI1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Llama-Song-Stream-3B-InstructIQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Llama-3.2-3B-Instruct-roleplay-tunedI1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Llama-3.2-3B-Instruct-heretic-ablitered-uncensoredI1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Llama3.2-3B-creative-writer-v0.1I1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Firefly-V3.2I1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Firefly-V3I1-IQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Hermes-3-Llama-3.2-3BIQ2_M3.2B1.14 GiB1.75 GiB3.70 GiB0.02 GiB23±30%
Yi-6B-ChatI1-IQ2_S6.1B1.88 GiB1.00 GiB3.70 GiB0.02 GiB23±30%
Qwen3.5-4B-NSFW-ARA-Heretic-LiteroticaI1-Q4_K_S4.2B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Qwen3.5-4B-RpRMax-v1I1-Q4_K_S4.7B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Holo-3.1-4B-uncensored-hereticI1-Q4_K_S4.5B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
GRaPE-2-MiniI1-Q4_K_S4.7B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Qwen3.5-DPO-4B-2I1-Q4_K_S4.2B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Qwen3.5-4B-BaseQ4_K_S4.7B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Huihui-Qwen3.5-4B-Claude-4.6-Opus-abliteratedI1-Q4_K_S4.7B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Qwopus3.5-4B-v3-hereticI1-Q4_K_S4.5B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
Aureth-4B-Qwen3.5I1-Q4_K_S4.5B2.39 GiB0.50 GiB3.70 GiB0.02 GiB23±30%
AMD-OLMo-1B-SFT-DPOQ6_K1.2B0.90 GiB2.00 GiB3.70 GiB0.02 GiB23±30%
MiniCPM-V-4Q5_K_M4.1B2.39 GiB0.50 GiB3.70 GiB0.02 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?
513 of 2118 indexed open-weight models fit a Arc A310 4GB at 16,384 context with f16 KV cache, the largest being Qwen3-Zero-Coder-Reasoning-V2-0.8B at I1-IQ2_M. 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.
Arc A310 4GB — what AI models can it run locally? — ossmodeldb