L
localai.computer
ModelsGPUsSystemsAI SetupsBuildsMethodology

Resources

  • Methodology
  • Submit Benchmark
  • About

Browse

  • AI Models
  • GPUs
  • PC Builds

Community

  • Leaderboard

Legal

  • Privacy
  • Terms
  • Contact

© 2025 localai.computer. Hardware recommendations for running AI models locally.

ℹ️We earn from qualifying purchases through affiliate links at no extra cost to you. This supports our free content and research.

  1. Home
  2. GPUs
  3. RX 7700 XT

Quick Answer: RX 7700 XT offers 12GB VRAM and starts around $399.99. It delivers approximately 90 tokens/sec on deepseek-ai/DeepSeek-OCR. It typically draws 245W under load.

RX 7700 XT

Unknown
By AMDReleased 2023-09MSRP $449.00

This GPU offers reliable throughput for local AI workloads. Pair it with the right model quantization to hit your desired tokens/sec, and monitor prices below to catch the best deal.

Buy on Amazon - $399.99View Benchmarks
Specs snapshot
Key hardware metrics for AI workloads.
VRAM12GB
Cores3,456
TDP245W
ArchitectureRDNA 3

Where to Buy

Buy directly on Amazon with fast shipping and reliable customer service.

AmazonUnknown
$399.99
Buy on Amazon

More Amazon options

Rotate out primary variants whenever validation flags an issue.

💡 Not ready to buy? Try cloud GPUs first

Test RX 7700 XT performance in the cloud before investing in hardware. Pay by the hour with no commitment.

Vast.aifrom $0.20/hrRunPodfrom $0.30/hrLambda Labsenterprise-grade

AI benchmarks

ModelQuantizationTokens/secVRAM used
deepseek-ai/DeepSeek-OCRQ4
89.90 tok/sEstimated

Auto-generated benchmark

2GB
ibm-research/PowerMoE-3bQ4
87.60 tok/sEstimated

Auto-generated benchmark

2GB
google-t5/t5-3bQ4
87.46 tok/sEstimated

Auto-generated benchmark

2GB
Qwen/Qwen2.5-3BQ4
87.31 tok/sEstimated

Auto-generated benchmark

2GB
ibm-granite/granite-3.3-2b-instructQ4
86.63 tok/sEstimated

Auto-generated benchmark

1GB
meta-llama/Llama-3.2-3BQ4
86.55 tok/sEstimated

Auto-generated benchmark

2GB
facebook/sam3Q4
85.73 tok/sEstimated

Auto-generated benchmark

1GB
google/gemma-3-1b-itQ4
85.24 tok/sEstimated

Auto-generated benchmark

1GB
unsloth/Llama-3.2-1B-InstructQ4
84.38 tok/sEstimated

Auto-generated benchmark

1GB
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16Q4
84.35 tok/sEstimated

Auto-generated benchmark

2GB
allenai/OLMo-2-0425-1BQ4
83.68 tok/sEstimated

Auto-generated benchmark

1GB
meta-llama/Llama-3.2-1B-InstructQ4
83.65 tok/sEstimated

Auto-generated benchmark

1GB
google/gemma-2-2b-itQ4
82.36 tok/sEstimated

Auto-generated benchmark

1GB
meta-llama/Llama-3.2-3B-InstructQ4
82.29 tok/sEstimated

Auto-generated benchmark

2GB
apple/OpenELM-1_1B-InstructQ4
82.19 tok/sEstimated

Auto-generated benchmark

1GB
deepseek-ai/deepseek-coder-1.3b-instructQ4
81.98 tok/sEstimated

Auto-generated benchmark

2GB
nari-labs/Dia2-2BQ4
80.86 tok/sEstimated

Auto-generated benchmark

2GB
Qwen/Qwen2.5-3B-InstructQ4
79.75 tok/sEstimated

Auto-generated benchmark

2GB
google-bert/bert-base-uncasedQ4
79.64 tok/sEstimated

Auto-generated benchmark

1GB
meta-llama/Llama-Guard-3-1BQ4
78.44 tok/sEstimated

Auto-generated benchmark

1GB
unsloth/Llama-3.2-3B-InstructQ4
78.13 tok/sEstimated

Auto-generated benchmark

2GB
LiquidAI/LFM2-1.2BQ4
77.50 tok/sEstimated

Auto-generated benchmark

1GB
inference-net/Schematron-3BQ4
77.24 tok/sEstimated

Auto-generated benchmark

2GB
unsloth/gemma-3-1b-itQ4
76.25 tok/sEstimated

Auto-generated benchmark

1GB
TinyLlama/TinyLlama-1.1B-Chat-v1.0Q4
75.82 tok/sEstimated

Auto-generated benchmark

1GB
meta-llama/Llama-3.2-1BQ4
75.51 tok/sEstimated

Auto-generated benchmark

1GB
tencent/HunyuanOCRQ4
75.50 tok/sEstimated

Auto-generated benchmark

1GB
bigcode/starcoder2-3bQ4
75.29 tok/sEstimated

Auto-generated benchmark

2GB
WeiboAI/VibeThinker-1.5BQ4
74.86 tok/sEstimated

Auto-generated benchmark

1GB
google/gemma-2bQ4
74.50 tok/sEstimated

Auto-generated benchmark

1GB
google/embeddinggemma-300mQ4
73.98 tok/sEstimated

Auto-generated benchmark

1GB
Qwen/Qwen3-8B-FP8Q4
73.96 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen3-0.6B-BaseQ4
73.80 tok/sEstimated

Auto-generated benchmark

3GB
Qwen/Qwen3-1.7BQ4
73.76 tok/sEstimated

Auto-generated benchmark

4GB
microsoft/Phi-3.5-vision-instructQ4
73.64 tok/sEstimated

Auto-generated benchmark

4GB
HuggingFaceH4/zephyr-7b-betaQ4
73.62 tok/sEstimated

Auto-generated benchmark

4GB
deepseek-ai/DeepSeek-R1Q4
73.33 tok/sEstimated

Auto-generated benchmark

4GB
vikhyatk/moondream2Q4
73.11 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen2.5-7B-InstructQ4
73.02 tok/sEstimated

Auto-generated benchmark

4GB
numind/NuExtract-1.5Q4
72.80 tok/sEstimated

Auto-generated benchmark

4GB
swiss-ai/Apertus-8B-Instruct-2509Q4
72.70 tok/sEstimated

Auto-generated benchmark

4GB
skt/kogpt2-base-v2Q4
72.64 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen2-7B-InstructQ4
72.63 tok/sEstimated

Auto-generated benchmark

4GB
GSAI-ML/LLaDA-8B-BaseQ4
72.54 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen2-0.5B-InstructQ4
72.53 tok/sEstimated

Auto-generated benchmark

3GB
openai-community/gpt2Q4
72.35 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen3-8B-BaseQ4
72.33 tok/sEstimated

Auto-generated benchmark

4GB
trl-internal-testing/tiny-LlamaForCausalLM-3.2Q4
72.29 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen-Image-Edit-2509Q4
72.08 tok/sEstimated

Auto-generated benchmark

4GB
Qwen/Qwen3-Reranker-0.6BQ4
71.73 tok/sEstimated

Auto-generated benchmark

3GB
deepseek-ai/DeepSeek-OCR
Q4
2GB
89.90 tok/sEstimated
Auto-generated benchmark
ibm-research/PowerMoE-3b
Q4
2GB
87.60 tok/sEstimated
Auto-generated benchmark
google-t5/t5-3b
Q4
2GB
87.46 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen2.5-3B
Q4
2GB
87.31 tok/sEstimated
Auto-generated benchmark
ibm-granite/granite-3.3-2b-instruct
Q4
1GB
86.63 tok/sEstimated
Auto-generated benchmark
meta-llama/Llama-3.2-3B
Q4
2GB
86.55 tok/sEstimated
Auto-generated benchmark
facebook/sam3
Q4
1GB
85.73 tok/sEstimated
Auto-generated benchmark
google/gemma-3-1b-it
Q4
1GB
85.24 tok/sEstimated
Auto-generated benchmark
unsloth/Llama-3.2-1B-Instruct
Q4
1GB
84.38 tok/sEstimated
Auto-generated benchmark
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16
Q4
2GB
84.35 tok/sEstimated
Auto-generated benchmark
allenai/OLMo-2-0425-1B
Q4
1GB
83.68 tok/sEstimated
Auto-generated benchmark
meta-llama/Llama-3.2-1B-Instruct
Q4
1GB
83.65 tok/sEstimated
Auto-generated benchmark
google/gemma-2-2b-it
Q4
1GB
82.36 tok/sEstimated
Auto-generated benchmark
meta-llama/Llama-3.2-3B-Instruct
Q4
2GB
82.29 tok/sEstimated
Auto-generated benchmark
apple/OpenELM-1_1B-Instruct
Q4
1GB
82.19 tok/sEstimated
Auto-generated benchmark
deepseek-ai/deepseek-coder-1.3b-instruct
Q4
2GB
81.98 tok/sEstimated
Auto-generated benchmark
nari-labs/Dia2-2B
Q4
2GB
80.86 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen2.5-3B-Instruct
Q4
2GB
79.75 tok/sEstimated
Auto-generated benchmark
google-bert/bert-base-uncased
Q4
1GB
79.64 tok/sEstimated
Auto-generated benchmark
meta-llama/Llama-Guard-3-1B
Q4
1GB
78.44 tok/sEstimated
Auto-generated benchmark
unsloth/Llama-3.2-3B-Instruct
Q4
2GB
78.13 tok/sEstimated
Auto-generated benchmark
LiquidAI/LFM2-1.2B
Q4
1GB
77.50 tok/sEstimated
Auto-generated benchmark
inference-net/Schematron-3B
Q4
2GB
77.24 tok/sEstimated
Auto-generated benchmark
unsloth/gemma-3-1b-it
Q4
1GB
76.25 tok/sEstimated
Auto-generated benchmark
TinyLlama/TinyLlama-1.1B-Chat-v1.0
Q4
1GB
75.82 tok/sEstimated
Auto-generated benchmark
meta-llama/Llama-3.2-1B
Q4
1GB
75.51 tok/sEstimated
Auto-generated benchmark
tencent/HunyuanOCR
Q4
1GB
75.50 tok/sEstimated
Auto-generated benchmark
bigcode/starcoder2-3b
Q4
2GB
75.29 tok/sEstimated
Auto-generated benchmark
WeiboAI/VibeThinker-1.5B
Q4
1GB
74.86 tok/sEstimated
Auto-generated benchmark
google/gemma-2b
Q4
1GB
74.50 tok/sEstimated
Auto-generated benchmark
google/embeddinggemma-300m
Q4
1GB
73.98 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen3-8B-FP8
Q4
4GB
73.96 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen3-0.6B-Base
Q4
3GB
73.80 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen3-1.7B
Q4
4GB
73.76 tok/sEstimated
Auto-generated benchmark
microsoft/Phi-3.5-vision-instruct
Q4
4GB
73.64 tok/sEstimated
Auto-generated benchmark
HuggingFaceH4/zephyr-7b-beta
Q4
4GB
73.62 tok/sEstimated
Auto-generated benchmark
deepseek-ai/DeepSeek-R1
Q4
4GB
73.33 tok/sEstimated
Auto-generated benchmark
vikhyatk/moondream2
Q4
4GB
73.11 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen2.5-7B-Instruct
Q4
4GB
73.02 tok/sEstimated
Auto-generated benchmark
numind/NuExtract-1.5
Q4
4GB
72.80 tok/sEstimated
Auto-generated benchmark
swiss-ai/Apertus-8B-Instruct-2509
Q4
4GB
72.70 tok/sEstimated
Auto-generated benchmark
skt/kogpt2-base-v2
Q4
4GB
72.64 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen2-7B-Instruct
Q4
4GB
72.63 tok/sEstimated
Auto-generated benchmark
GSAI-ML/LLaDA-8B-Base
Q4
4GB
72.54 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen2-0.5B-Instruct
Q4
3GB
72.53 tok/sEstimated
Auto-generated benchmark
openai-community/gpt2
Q4
4GB
72.35 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen3-8B-Base
Q4
4GB
72.33 tok/sEstimated
Auto-generated benchmark
trl-internal-testing/tiny-LlamaForCausalLM-3.2
Q4
4GB
72.29 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen-Image-Edit-2509
Q4
4GB
72.08 tok/sEstimated
Auto-generated benchmark
Qwen/Qwen3-Reranker-0.6B
Q4
3GB
71.73 tok/sEstimated
Auto-generated benchmark

Note: Performance estimates are calculated. Real results may vary. Methodology · Submit real data

Model compatibility

ModelQuantizationVerdictEstimated speedVRAM needed
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16Q4Fits comfortably
84.35 tok/sEstimated
2GB (have 12GB)
distilbert/distilgpt2Q4Fits comfortably
71.00 tok/sEstimated
4GB (have 12GB)
distilbert/distilgpt2Q8Fits comfortably
44.55 tok/sEstimated
7GB (have 12GB)
meta-llama/Meta-Llama-3-8B-InstructQ4Fits comfortably
69.70 tok/sEstimated
4GB (have 12GB)
meta-llama/Meta-Llama-3-8B-InstructQ8Fits comfortably
42.57 tok/sEstimated
9GB (have 12GB)
meta-llama/Meta-Llama-3-8B-InstructFP16Not supported
23.25 tok/sEstimated
17GB (have 12GB)
Qwen/Qwen2.5-1.5BQ4Fits comfortably
64.14 tok/sEstimated
3GB (have 12GB)
Qwen/Qwen2.5-1.5BQ8Fits comfortably
45.67 tok/sEstimated
5GB (have 12GB)
unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bitQ4Fits comfortably
69.59 tok/sEstimated
4GB (have 12GB)
zai-org/GLM-4.6-FP8Q8Fits comfortably
44.46 tok/sEstimated
7GB (have 12GB)
Qwen/Qwen2.5-32B-InstructQ8Not supported
15.36 tok/sEstimated
33GB (have 12GB)
Qwen/Qwen2.5-32B-InstructFP16Not supported
8.22 tok/sEstimated
66GB (have 12GB)
mistralai/Mistral-7B-v0.1Q4Fits comfortably
62.48 tok/sEstimated
4GB (have 12GB)
HuggingFaceTB/SmolLM-135MQ8Fits comfortably
48.75 tok/sEstimated
7GB (have 12GB)
HuggingFaceTB/SmolLM-135MFP16Not supported
23.14 tok/sEstimated
15GB (have 12GB)
trl-internal-testing/tiny-random-LlamaForCausalLMFP16Not supported
23.73 tok/sEstimated
15GB (have 12GB)
Qwen/Qwen3-14B-BaseQ8Not supported
37.35 tok/sEstimated
14GB (have 12GB)
Qwen/Qwen3-14B-BaseFP16Not supported
20.49 tok/sEstimated
29GB (have 12GB)
meta-llama/Llama-2-7b-chat-hfQ4Fits comfortably
65.29 tok/sEstimated
4GB (have 12GB)
unsloth/DeepSeek-R1-Distill-Qwen-32B-bnb-4bitFP16Not supported
8.39 tok/sEstimated
66GB (have 12GB)
RedHatAI/Meta-Llama-3.1-70B-Instruct-quantized.w4a16FP16Not supported
9.07 tok/sEstimated
137GB (have 12GB)
OpenPipe/Qwen3-14B-InstructQ4Fits comfortably
49.80 tok/sEstimated
7GB (have 12GB)
numind/NuExtract-1.5FP16Not supported
25.51 tok/sEstimated
15GB (have 12GB)
Qwen/Qwen2.5-Coder-7B-InstructQ4Fits comfortably
70.25 tok/sEstimated
4GB (have 12GB)
Qwen/Qwen2.5-Coder-7B-InstructFP16Not supported
27.31 tok/sEstimated
15GB (have 12GB)
Qwen/Qwen3-4B-Thinking-2507Q4Fits comfortably
63.43 tok/sEstimated
2GB (have 12GB)
ibm-granite/granite-docling-258MQ8Fits comfortably
45.40 tok/sEstimated
7GB (have 12GB)
ibm-granite/granite-docling-258MFP16Not supported
26.45 tok/sEstimated
15GB (have 12GB)
bigcode/starcoder2-3bFP16Fits comfortably
31.91 tok/sEstimated
6GB (have 12GB)
unsloth/gemma-3-1b-itQ8Fits comfortably
53.13 tok/sEstimated
1GB (have 12GB)
unsloth/gemma-3-1b-itFP16Fits comfortably
33.37 tok/sEstimated
2GB (have 12GB)
nvidia/NVIDIA-Nemotron-Nano-9B-v2FP16Not supported
20.72 tok/sEstimated
19GB (have 12GB)
openai/gpt-oss-safeguard-20bQ8Not supported
26.07 tok/sEstimated
22GB (have 12GB)
openai/gpt-oss-safeguard-20bFP16Not supported
15.13 tok/sEstimated
44GB (have 12GB)
MiniMaxAI/MiniMax-M1-40kQ4Not supported
8.88 tok/sEstimated
255GB (have 12GB)
MiniMaxAI/MiniMax-M1-40kQ8Not supported
6.21 tok/sEstimated
510GB (have 12GB)
mistralai/Mistral-Large-3-675B-Instruct-2512Q4Not supported
7.46 tok/sEstimated
378GB (have 12GB)
mistralai/Mistral-Large-3-675B-Instruct-2512Q8Not supported
5.43 tok/sEstimated
755GB (have 12GB)
mistralai/Mistral-Large-3-675B-Instruct-2512FP16Not supported
2.86 tok/sEstimated
1509GB (have 12GB)
EssentialAI/rnj-1Q4Fits comfortably
49.63 tok/sEstimated
5GB (have 12GB)
EssentialAI/rnj-1Q8Fits comfortably
38.56 tok/sEstimated
10GB (have 12GB)
EssentialAI/rnj-1FP16Not supported
19.77 tok/sEstimated
19GB (have 12GB)
Qwen/Qwen2.5-3B-InstructFP16Fits comfortably
33.74 tok/sEstimated
6GB (have 12GB)
bigscience/bloomz-560mQ4Fits comfortably
67.53 tok/sEstimated
4GB (have 12GB)
bigscience/bloomz-560mQ8Fits comfortably
48.82 tok/sEstimated
7GB (have 12GB)
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16FP16Fits comfortably
30.24 tok/sEstimated
6GB (have 12GB)
mistralai/Mistral-7B-Instruct-v0.2Q4Fits comfortably
63.53 tok/sEstimated
4GB (have 12GB)
mistralai/Mistral-7B-Instruct-v0.2Q8Fits comfortably
50.28 tok/sEstimated
7GB (have 12GB)
mistralai/Mistral-7B-Instruct-v0.2FP16Not supported
26.66 tok/sEstimated
15GB (have 12GB)
bigscience/bloomz-560mFP16Not supported
27.40 tok/sEstimated
15GB (have 12GB)
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16Q4
Fits comfortably2GB required · 12GB available
84.35 tok/sEstimated
distilbert/distilgpt2Q4
Fits comfortably4GB required · 12GB available
71.00 tok/sEstimated
distilbert/distilgpt2Q8
Fits comfortably7GB required · 12GB available
44.55 tok/sEstimated
meta-llama/Meta-Llama-3-8B-InstructQ4
Fits comfortably4GB required · 12GB available
69.70 tok/sEstimated
meta-llama/Meta-Llama-3-8B-InstructQ8
Fits comfortably9GB required · 12GB available
42.57 tok/sEstimated
meta-llama/Meta-Llama-3-8B-InstructFP16
Not supported17GB required · 12GB available
23.25 tok/sEstimated
Qwen/Qwen2.5-1.5BQ4
Fits comfortably3GB required · 12GB available
64.14 tok/sEstimated
Qwen/Qwen2.5-1.5BQ8
Fits comfortably5GB required · 12GB available
45.67 tok/sEstimated
unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bitQ4
Fits comfortably4GB required · 12GB available
69.59 tok/sEstimated
zai-org/GLM-4.6-FP8Q8
Fits comfortably7GB required · 12GB available
44.46 tok/sEstimated
Qwen/Qwen2.5-32B-InstructQ8
Not supported33GB required · 12GB available
15.36 tok/sEstimated
Qwen/Qwen2.5-32B-InstructFP16
Not supported66GB required · 12GB available
8.22 tok/sEstimated
mistralai/Mistral-7B-v0.1Q4
Fits comfortably4GB required · 12GB available
62.48 tok/sEstimated
HuggingFaceTB/SmolLM-135MQ8
Fits comfortably7GB required · 12GB available
48.75 tok/sEstimated
HuggingFaceTB/SmolLM-135MFP16
Not supported15GB required · 12GB available
23.14 tok/sEstimated
trl-internal-testing/tiny-random-LlamaForCausalLMFP16
Not supported15GB required · 12GB available
23.73 tok/sEstimated
Qwen/Qwen3-14B-BaseQ8
Not supported14GB required · 12GB available
37.35 tok/sEstimated
Qwen/Qwen3-14B-BaseFP16
Not supported29GB required · 12GB available
20.49 tok/sEstimated
meta-llama/Llama-2-7b-chat-hfQ4
Fits comfortably4GB required · 12GB available
65.29 tok/sEstimated
unsloth/DeepSeek-R1-Distill-Qwen-32B-bnb-4bitFP16
Not supported66GB required · 12GB available
8.39 tok/sEstimated
RedHatAI/Meta-Llama-3.1-70B-Instruct-quantized.w4a16FP16
Not supported137GB required · 12GB available
9.07 tok/sEstimated
OpenPipe/Qwen3-14B-InstructQ4
Fits comfortably7GB required · 12GB available
49.80 tok/sEstimated
numind/NuExtract-1.5FP16
Not supported15GB required · 12GB available
25.51 tok/sEstimated
Qwen/Qwen2.5-Coder-7B-InstructQ4
Fits comfortably4GB required · 12GB available
70.25 tok/sEstimated
Qwen/Qwen2.5-Coder-7B-InstructFP16
Not supported15GB required · 12GB available
27.31 tok/sEstimated
Qwen/Qwen3-4B-Thinking-2507Q4
Fits comfortably2GB required · 12GB available
63.43 tok/sEstimated
ibm-granite/granite-docling-258MQ8
Fits comfortably7GB required · 12GB available
45.40 tok/sEstimated
ibm-granite/granite-docling-258MFP16
Not supported15GB required · 12GB available
26.45 tok/sEstimated
bigcode/starcoder2-3bFP16
Fits comfortably6GB required · 12GB available
31.91 tok/sEstimated
unsloth/gemma-3-1b-itQ8
Fits comfortably1GB required · 12GB available
53.13 tok/sEstimated
unsloth/gemma-3-1b-itFP16
Fits comfortably2GB required · 12GB available
33.37 tok/sEstimated
nvidia/NVIDIA-Nemotron-Nano-9B-v2FP16
Not supported19GB required · 12GB available
20.72 tok/sEstimated
openai/gpt-oss-safeguard-20bQ8
Not supported22GB required · 12GB available
26.07 tok/sEstimated
openai/gpt-oss-safeguard-20bFP16
Not supported44GB required · 12GB available
15.13 tok/sEstimated
MiniMaxAI/MiniMax-M1-40kQ4
Not supported255GB required · 12GB available
8.88 tok/sEstimated
MiniMaxAI/MiniMax-M1-40kQ8
Not supported510GB required · 12GB available
6.21 tok/sEstimated
mistralai/Mistral-Large-3-675B-Instruct-2512Q4
Not supported378GB required · 12GB available
7.46 tok/sEstimated
mistralai/Mistral-Large-3-675B-Instruct-2512Q8
Not supported755GB required · 12GB available
5.43 tok/sEstimated
mistralai/Mistral-Large-3-675B-Instruct-2512FP16
Not supported1509GB required · 12GB available
2.86 tok/sEstimated
EssentialAI/rnj-1Q4
Fits comfortably5GB required · 12GB available
49.63 tok/sEstimated
EssentialAI/rnj-1Q8
Fits comfortably10GB required · 12GB available
38.56 tok/sEstimated
EssentialAI/rnj-1FP16
Not supported19GB required · 12GB available
19.77 tok/sEstimated
Qwen/Qwen2.5-3B-InstructFP16
Fits comfortably6GB required · 12GB available
33.74 tok/sEstimated
bigscience/bloomz-560mQ4
Fits comfortably4GB required · 12GB available
67.53 tok/sEstimated
bigscience/bloomz-560mQ8
Fits comfortably7GB required · 12GB available
48.82 tok/sEstimated
context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16FP16
Fits comfortably6GB required · 12GB available
30.24 tok/sEstimated
mistralai/Mistral-7B-Instruct-v0.2Q4
Fits comfortably4GB required · 12GB available
63.53 tok/sEstimated
mistralai/Mistral-7B-Instruct-v0.2Q8
Fits comfortably7GB required · 12GB available
50.28 tok/sEstimated
mistralai/Mistral-7B-Instruct-v0.2FP16
Not supported15GB required · 12GB available
26.66 tok/sEstimated
bigscience/bloomz-560mFP16
Not supported15GB required · 12GB available
27.40 tok/sEstimated

Note: Performance estimates are calculated. Real results may vary. Methodology · Submit real data

Alternative GPUs

RTX 5070
12GB

Explore how RTX 5070 stacks up for local inference workloads.

RTX 4060 Ti 16GB
16GB

Explore how RTX 4060 Ti 16GB stacks up for local inference workloads.

RX 6800 XT
16GB

Explore how RX 6800 XT stacks up for local inference workloads.

RTX 4070 Super
12GB

Explore how RTX 4070 Super stacks up for local inference workloads.

RTX 3080
10GB

Explore how RTX 3080 stacks up for local inference workloads.