NVIDIA · workstation

RTX A1000

RTX A1000 has 8 GB of VRAM at 192 GB/s — about 7.44 GiB usable after driver and compositor overhead. 1398 of 2118 indexed models fit at 8K context with q8_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
8 GB
GDDR6
Bandwidth
192 GB/s
128-bit bus
Tensor FP16
27 TF
dense
TDP
50 W
$365 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1203vision language 101embedding 26video 7audio tts 21image 2audio asr 38

What fits at 8K context

largest quantization that fits, per model · 1398 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
MiroThinker-v1.0-8BQ5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
Qwen3-8B-abliteratedQ5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
Qwen3-8BQ5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
Josiefied-Qwen3-8B-abliterated-v1Q5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
Nemotron-Orchestrator-8BQ5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
DeepSeek-R1-0528-Qwen3-8BQ5_K_L8.2B5.81 GiB0.60 GiB7.44 GiB0.00 GiB17±22%
Qwen3.5-9BQ5_K_M9.7B6.27 GiB0.13 GiB7.44 GiB0.00 GiB17±22%
AceReason-Nemotron-14BUD-IQ3_XXS14.8B5.59 GiB0.80 GiB7.44 GiB0.00 GiB17±22%
gemma-4-12BQ3_K_M12.0B5.87 GiB0.51 GiB7.43 GiB0.01 GiB17±22%
Grug-12BQ3_K_M12.0B5.87 GiB0.51 GiB7.43 GiB0.01 GiB17±22%
gemma-4-12B-it-Esper4Q3_K_M12.0B5.87 GiB0.51 GiB7.43 GiB0.01 GiB17±22%
gemma-4-12B-itQ3_K_M12.0B5.87 GiB0.51 GiB7.43 GiB0.01 GiB17±22%
Hunyuan-7B-InstructQ6_K_L7.5B5.86 GiB0.53 GiB7.43 GiB0.01 GiB17±22%
LFM2-8B-A1BMoEQ6_K8.3B6.38 GiB0.05 GiB7.42 GiB0.02 GiB52±37%
codegeex4-all-9bQ2_K9.4B3.72 GiB2.66 GiB7.42 GiB0.02 GiB17±22%
Qwen3.6-28BMoEI1-IQ1_M28.2B6.33 GiB0.08 GiB7.42 GiB0.02 GiB86±37%
Qwen3.5-28BMoEI1-IQ1_M28.7B6.33 GiB0.08 GiB7.42 GiB0.02 GiB86±37%
Llama3.2-24B-A3B-II-Dark-Champion-INSTRUCT-Heretic-Abliterated-UncensoredMoEI1-Q2_K_S18.0B5.95 GiB0.46 GiB7.42 GiB0.02 GiB41±37%
Qwen3-VL-Embedding-8BQ6_K8.1B5.79 GiB0.60 GiB7.42 GiB0.02 GiB17±22%
granite-3.3-8b-instructQ5_18.2B5.72 GiB0.66 GiB7.42 GiB0.02 GiB17±22%
glm-4-9b-chat-abliteratedQ2_K9.4B3.72 GiB2.66 GiB7.42 GiB0.02 GiB17±22%
glm-4-9b-chatQ2_K9.4B3.72 GiB2.66 GiB7.42 GiB0.02 GiB17±22%
granite-3.2-8b-instructQ5_18.2B5.72 GiB0.66 GiB7.42 GiB0.02 GiB17±22%
Qwen3-8B-BaseQ6_K8.2B5.79 GiB0.60 GiB7.42 GiB0.02 GiB17±22%
qwen-indic-v1I1-Q6_K7.6B5.79 GiB0.60 GiB7.42 GiB0.02 GiB17±22%
Qwen3-Embedding-8BQ6_K7.6B5.79 GiB0.60 GiB7.42 GiB0.02 GiB17±22%
Falcon3-7B-InstructQ6_K_L7.5B5.88 GiB0.46 GiB7.41 GiB0.03 GiB17±22%
v6-Finch-7B-HFIQ4_XS7.6B4.24 GiB2.13 GiB7.41 GiB0.03 GiB17±22%
rwkv-6-world-7bIQ4_XS7.6B4.24 GiB2.13 GiB7.41 GiB0.03 GiB17±22%
granite-3.1-8b-instructQ5_18.2B5.71 GiB0.66 GiB7.41 GiB0.03 GiB17±22%
Fimbulvetr-11B-v2IQ4_XS10.7B5.57 GiB0.80 GiB7.41 GiB0.03 GiB17±22%
NVIDIA-Nemotron-Nano-9B-v2Q4_18.9B5.43 GiB0.93 GiB7.40 GiB0.04 GiB17±22%
Ministral-3-14B-Instruct-2512-BF16-abliteratedI1-IQ3_S13.9B5.68 GiB0.66 GiB7.40 GiB0.04 GiB17±22%
Ministral-3-14B-Reasoning-2512-UncensoredI1-IQ3_S13.9B5.68 GiB0.66 GiB7.40 GiB0.04 GiB17±22%
EVA-Yi-1.5-9B-32K-V1Q5_K_L8.8B5.98 GiB0.40 GiB7.40 GiB0.04 GiB17±22%
Yi-Coder-9B-ChatQ5_K_L8.8B5.98 GiB0.40 GiB7.40 GiB0.04 GiB17±22%
Mistral-7B-v0.3Q6_K_L7.2B5.83 GiB0.53 GiB7.40 GiB0.04 GiB17±22%
Gemma-The-Writer-N-Restless-Quill-10B-UncensoredI1-Q3_K_L10.0B5.17 GiB1.19 GiB7.40 GiB0.04 GiB17±22%
NVIDIA-Nemotron-Nano-12B-v2IQ3_M12.3B5.30 GiB1.03 GiB7.40 GiB0.04 GiB17±22%
Maestro1-9BQ5_18.8B5.77 GiB0.60 GiB7.40 GiB0.04 GiB17±22%
Jan-v2-VL-highQ5_18.8B5.77 GiB0.60 GiB7.40 GiB0.04 GiB17±22%
Jan-v2-VL-medQ5_18.8B5.77 GiB0.60 GiB7.40 GiB0.04 GiB17±22%
MiniCPM-o-4_5Q5_19.4B5.77 GiB0.60 GiB7.40 GiB0.04 GiB17±22%
Wan2.2-Animate-14BQ2_K17.3B6.36 GiB0.00 GiB7.40 GiB0.04 GiB17±22%
Qwen3.6-14B-A3B-FableVibesMoEQ3_K_M13.8B6.30 GiB0.08 GiB7.39 GiB0.05 GiB62±37%
Qwen3.6-14B-A3B-VibeForged-v2MoEQ3_K_M13.8B6.30 GiB0.08 GiB7.39 GiB0.05 GiB62±37%
MN-GRAND-23.5B-Gutenberg-UNCENSORED-V2-GLM4.7-ThinkingI1-IQ1_S23.4B5.00 GiB1.34 GiB7.39 GiB0.05 GiB17±22%
glm-4-9b-chat-1mIQ2_M9.5B3.69 GiB2.66 GiB7.39 GiB0.05 GiB17±22%
Devstral-Small-2-24B-Instruct-2512UD-IQ1_M24.0B5.60 GiB0.66 GiB7.39 GiB0.05 GiB17±22%
Mistral-Small-3.2-24B-Instruct-2506UD-IQ1_M24.0B5.60 GiB0.66 GiB7.39 GiB0.05 GiB18±22%
Devstral-Small-2507UD-IQ1_M23.6B5.60 GiB0.66 GiB7.39 GiB0.05 GiB18±22%
Devstral-Small-2505UD-IQ1_M23.6B5.60 GiB0.66 GiB7.39 GiB0.05 GiB18±22%
Magistral-Small-2507UD-IQ1_M23.6B5.60 GiB0.66 GiB7.39 GiB0.05 GiB18±22%
Mistral-Small-3.1-24B-Instruct-2503UD-IQ1_M24.0B5.60 GiB0.66 GiB7.39 GiB0.05 GiB18±22%
EVA-abliterated-TIES-Qwen2.5-14BI1-IQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
Neuron-V1-14B-InstructI1-IQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
Ektome-Qwen2.5-Coder-14B-Instruct-PristinelyUncensoredI1-IQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
Qwen2.5-14B-Instruct-1M-abliteratedI1-IQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
DeepCoder-14B-PreviewIQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
Deepseeker-Kunou-Qwen2.5-14bI1-IQ3_XXS14.8B5.54 GiB0.80 GiB7.38 GiB0.06 GiB17±22%
●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.

Measured on this card

third-party benchmarks, aggregated
Workload◍MedianMiddle 50%Runs
Image generation3.75 it/s3.59–4.057
Benchmarked· n=7

Aggregated from community-submitted runs, so the spread is wide by nature — it covers different models, resolutions, step counts and settings, not one controlled configuration. Read the middle 50% rather than the median alone. These figures are reproduced with attribution from vladmandic-sd-data-benchmark, which publishes no licence — so we display and link rather than redistribute them.

Questions people ask

What AI models can a RTX A1000 run?
1398 of 2118 indexed open-weight models fit a RTX A1000 at 8,192 context with q8_0 KV cache, the largest being MiroThinker-v1.0-8B at Q5_K_L. That covers text, vision-language, image, video and speech models.
How much usable memory does a RTX A1000 actually have?
Its nameplate is 8 GB, but about 7.44 GiB is available to a model once driver and compositor overhead is accounted for.
Is a RTX A1000 fast for local AI?
Its memory bandwidth is 192 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.