Quick Answer: RTX 5070 offers 12GB VRAM and starts around $761.10. It delivers approximately 142 tokens/sec on Qwen/Qwen3-ASR-1.7B. It typically draws 250W under load.
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.
With 12GB VRAM, RTX 5070 can run models up to approximately 30B parameters using 4-bit quantization. This is suitable for 7B-13B models like Llama 3 8B, Mistral 7B, and Qwen 7B.
Consider RTX 4080 Super or RTX 4090 — More VRAM and cores for demanding workloads.
| Model | Quantization | Tokens/sec | VRAM used |
|---|---|---|---|
| Qwen/Qwen3-ASR-1.7B | Q4 | 141.72 tok/sEstimated Auto-generated benchmark | 2GB |
| google-bert/bert-base-uncased | Q4 | 140.01 tok/sEstimated Auto-generated benchmark | 1GB |
| ibm-granite/granite-3.3-2b-instruct | Q4 | 138.92 tok/sEstimated Auto-generated benchmark | 1GB |
| meta-llama/Llama-3.2-3B | Q4 | 137.40 tok/sEstimated Auto-generated benchmark | 2GB |
| ibm-research/PowerMoE-3b | Q4 | 136.36 tok/sEstimated Auto-generated benchmark | 2GB |
| deepseek-ai/DeepSeek-OCR | Q4 | 135.92 tok/sEstimated Auto-generated benchmark | 2GB |
| deepseek-ai/deepseek-coder-1.3b-instruct | Q4 | 135.19 tok/sEstimated Auto-generated benchmark | 2GB |
| inference-net/Schematron-3B | Q4 | 134.90 tok/sEstimated Auto-generated benchmark | 2GB |
| google/gemma-2b | Q4 | 134.86 tok/sEstimated Auto-generated benchmark | 1GB |
| google/gemma-3-1b-it | Q4 | 134.66 tok/sEstimated Auto-generated benchmark | 1GB |
| context-labs/meta-llama-Llama-3.2-3B-Instruct-FP16 | Q4 | 133.94 tok/sEstimated Auto-generated benchmark | 2GB |
| facebook/sam3 | Q4 | 133.32 tok/sEstimated Auto-generated benchmark | 1GB |
| Qwen/Qwen2.5-3B-Instruct | Q4 | 133.21 tok/sEstimated Auto-generated benchmark | 2GB |
| bigcode/starcoder2-3b | Q4 | 132.96 tok/sEstimated Auto-generated benchmark | 2GB |
| unsloth/gemma-3-1b-it | Q4 | 132.93 tok/sEstimated Auto-generated benchmark | 1GB |
| TinyLlama/TinyLlama-1.1B-Chat-v1.0 | Q4 | 132.92 tok/sEstimated Auto-generated benchmark | 1GB |
| deepseek-ai/DeepSeek-OCR-2 | Q4 | 132.19 tok/sEstimated Auto-generated benchmark | 2GB |
| google-t5/t5-3b | Q4 | 130.41 tok/sEstimated Auto-generated benchmark | 2GB |
| nari-labs/Dia2-2B | Q4 | 128.99 tok/sEstimated Auto-generated benchmark | 2GB |
| google/gemma-2-2b-it | Q4 | 128.98 tok/sEstimated Auto-generated benchmark | 1GB |
| unsloth/Llama-3.2-1B-Instruct | Q4 | 126.30 tok/sEstimated Auto-generated benchmark | 1GB |
| meta-llama/Llama-3.2-3B-Instruct | Q4 | 126.09 tok/sEstimated Auto-generated benchmark | 2GB |
| meta-llama/Llama-3.2-1B | Q4 | 125.75 tok/sEstimated Auto-generated benchmark | 1GB |
| nineninesix/kani-tts-2-en | Q4 | 125.46 tok/sEstimated Auto-generated benchmark | 1GB |
| LiquidAI/LFM2-1.2B | Q4 | 124.86 tok/sEstimated Auto-generated benchmark | 1GB |
| meta-llama/Llama-Guard-3-1B | Q4 | 123.98 tok/sEstimated Auto-generated benchmark | 1GB |
| apple/OpenELM-1_1B-Instruct | Q4 | 123.74 tok/sEstimated Auto-generated benchmark | 1GB |
| tencent/HunyuanOCR | Q4 | 123.41 tok/sEstimated Auto-generated benchmark | 1GB |
| meta-llama/Llama-3.2-1B-Instruct | Q4 | 122.79 tok/sEstimated Auto-generated benchmark | 1GB |
| unsloth/Llama-3.2-3B-Instruct | Q4 | 122.20 tok/sEstimated Auto-generated benchmark | 2GB |
| WeiboAI/VibeThinker-1.5B | Q4 | 121.39 tok/sEstimated Auto-generated benchmark | 1GB |
| allenai/OLMo-2-0425-1B | Q4 | 120.48 tok/sEstimated Auto-generated benchmark | 1GB |
| google/embeddinggemma-300m | Q4 | 119.38 tok/sEstimated Auto-generated benchmark | 1GB |
| Qwen/Qwen2.5-3B | Q4 | 118.90 tok/sEstimated Auto-generated benchmark | 2GB |
| swiss-ai/Apertus-8B-Instruct-2509 | Q4 | 118.29 tok/sEstimated Auto-generated benchmark | 4GB |
| microsoft/Phi-3-mini-128k-instruct | Q4 | 118.22 tok/sEstimated Auto-generated benchmark | 4GB |
| Qwen/Qwen2.5-0.5B | Q4 | 118.02 tok/sEstimated Auto-generated benchmark | 3GB |
| huggyllama/llama-7b | Q4 | 117.65 tok/sEstimated Auto-generated benchmark | 4GB |
| FireRedTeam/FireRed-Image-Edit-1.0 | Q4 | 117.33 tok/sEstimated Auto-generated benchmark | 4GB |
| lmstudio-community/DeepSeek-R1-0528-Qwen3-8B-MLX-8bit | Q4 | 117.29 tok/sEstimated Auto-generated benchmark | 4GB |
| meta-llama/Meta-Llama-3-8B | Q4 | 117.28 tok/sEstimated Auto-generated benchmark | 4GB |
| lmstudio-community/Qwen3-4B-Thinking-2507-MLX-8bit | Q4 | 117.18 tok/sEstimated Auto-generated benchmark | 2GB |
| Qwen/Qwen2.5-Coder-7B-Instruct | Q4 | 116.88 tok/sEstimated Auto-generated benchmark | 4GB |
| deepseek-ai/DeepSeek-V3-0324 | Q4 | 116.57 tok/sEstimated Auto-generated benchmark | 4GB |
| Qwen/Qwen3-Embedding-0.6B | Q4 | 116.52 tok/sEstimated Auto-generated benchmark | 3GB |
| GSAI-ML/LLaDA-8B-Base | Q4 | 116.41 tok/sEstimated Auto-generated benchmark | 4GB |
| meta-llama/Meta-Llama-3-8B-Instruct | Q4 | 116.11 tok/sEstimated Auto-generated benchmark | 4GB |
| microsoft/DialoGPT-medium | Q4 | 116.06 tok/sEstimated Auto-generated benchmark | 4GB |
| deepseek-ai/DeepSeek-R1 | Q4 | 115.89 tok/sEstimated Auto-generated benchmark | 4GB |
| FireRedTeam/FireRed-Image-Edit-1.0 | Q4 | 115.68 tok/sEstimated Auto-generated benchmark | 4GB |
Note: Performance estimates are calculated. Real results may vary. Methodology · Submit real data
| Model | Quantization | Verdict | Estimated speed | VRAM needed |
|---|---|---|---|---|
| openai-community/gpt2 | Q8 | Fits comfortably | 73.80 tok/sEstimated | 7GB (have 12GB) |
| openai-community/gpt2 | FP16 | Not supported | 44.28 tok/sEstimated | 15GB (have 12GB) |
| Qwen/Qwen2.5-7B-Instruct | Q4 | Fits comfortably | 105.45 tok/sEstimated | 4GB (have 12GB) |
| Qwen/Qwen2.5-7B-Instruct | Q8 | Fits comfortably | 77.89 tok/sEstimated | 7GB (have 12GB) |
| Qwen/Qwen2.5-7B-Instruct | FP16 | Not supported | 41.72 tok/sEstimated | 15GB (have 12GB) |
| Qwen/Qwen3-0.6B | Q4 | Fits comfortably | 105.76 tok/sEstimated | 3GB (have 12GB) |
| Qwen/Qwen3-0.6B | Q8 | Fits comfortably | 80.08 tok/sEstimated | 6GB (have 12GB) |
| Qwen/Qwen3-0.6B | FP16 | Not supported | 40.66 tok/sEstimated | 13GB (have 12GB) |
| Gensyn/Qwen2.5-0.5B-Instruct | Q4 | Fits comfortably | 113.19 tok/sEstimated | 3GB (have 12GB) |
| Gensyn/Qwen2.5-0.5B-Instruct | Q8 | Fits comfortably | 80.78 tok/sEstimated | 5GB (have 12GB) |
| Gensyn/Qwen2.5-0.5B-Instruct | FP16 | Fits (tight) | 42.28 tok/sEstimated | 11GB (have 12GB) |
| meta-llama/Llama-3.1-8B-Instruct | Q4 | Fits comfortably | 105.69 tok/sEstimated | 4GB (have 12GB) |
| meta-llama/Llama-3.1-8B-Instruct | Q8 | Fits comfortably | 77.38 tok/sEstimated | 9GB (have 12GB) |
| meta-llama/Llama-3.1-8B-Instruct | FP16 | Not supported | 44.35 tok/sEstimated | 17GB (have 12GB) |
| dphn/dolphin-2.9.1-yi-1.5-34b | Q4 | Not supported | 35.11 tok/sEstimated | 17GB (have 12GB) |
| dphn/dolphin-2.9.1-yi-1.5-34b | Q8 | Not supported | 24.74 tok/sEstimated | 35GB (have 12GB) |
| dphn/dolphin-2.9.1-yi-1.5-34b | FP16 | Not supported | 15.67 tok/sEstimated | 70GB (have 12GB) |
| openai/gpt-oss-20b | Q4 | Fits comfortably | 60.23 tok/sEstimated | 10GB (have 12GB) |
| openai/gpt-oss-20b | Q8 | Not supported | 42.10 tok/sEstimated | 20GB (have 12GB) |
| openai/gpt-oss-20b | FP16 | Not supported | 21.14 tok/sEstimated | 41GB (have 12GB) |
| google/gemma-3-1b-it | Q4 | Fits comfortably | 134.66 tok/sEstimated | 1GB (have 12GB) |
| google/gemma-3-1b-it | Q8 | Fits comfortably | 94.76 tok/sEstimated | 1GB (have 12GB) |
| google/gemma-3-1b-it | FP16 | Fits comfortably | 45.16 tok/sEstimated | 2GB (have 12GB) |
| Qwen/Qwen3-Embedding-0.6B | Q4 | Fits comfortably | 116.52 tok/sEstimated | 3GB (have 12GB) |
| Qwen/Qwen3-Embedding-0.6B | Q8 | Fits comfortably | 79.85 tok/sEstimated | 6GB (have 12GB) |
| Qwen/Qwen3-Embedding-0.6B | FP16 | Not supported | 44.55 tok/sEstimated | 13GB (have 12GB) |
| Qwen/Qwen2.5-1.5B-Instruct | Q4 | Fits comfortably | 107.28 tok/sEstimated | 3GB (have 12GB) |
| Qwen/Qwen2.5-1.5B-Instruct | Q8 | Fits comfortably | 69.54 tok/sEstimated | 5GB (have 12GB) |
| Qwen/Qwen2.5-1.5B-Instruct | FP16 | Fits (tight) | 38.90 tok/sEstimated | 11GB (have 12GB) |
| facebook/opt-125m | Q4 | Fits comfortably | 99.36 tok/sEstimated | 4GB (have 12GB) |
| facebook/opt-125m | Q8 | Fits comfortably | 79.73 tok/sEstimated | 7GB (have 12GB) |
| facebook/opt-125m | FP16 | Not supported | 43.52 tok/sEstimated | 15GB (have 12GB) |
| TinyLlama/TinyLlama-1.1B-Chat-v1.0 | Q4 | Fits comfortably | 132.92 tok/sEstimated | 1GB (have 12GB) |
| TinyLlama/TinyLlama-1.1B-Chat-v1.0 | Q8 | Fits comfortably | 97.71 tok/sEstimated | 1GB (have 12GB) |
| TinyLlama/TinyLlama-1.1B-Chat-v1.0 | FP16 | Fits comfortably | 50.58 tok/sEstimated | 2GB (have 12GB) |
| trl-internal-testing/tiny-Qwen2ForCausalLM-2.5 | Q4 | Fits comfortably | 96.87 tok/sEstimated | 4GB (have 12GB) |
| trl-internal-testing/tiny-Qwen2ForCausalLM-2.5 | Q8 | Fits comfortably | 77.81 tok/sEstimated | 7GB (have 12GB) |
| trl-internal-testing/tiny-Qwen2ForCausalLM-2.5 | FP16 | Not supported | 38.40 tok/sEstimated | 15GB (have 12GB) |
| Qwen/Qwen3-4B-Instruct-2507 | Q4 | Fits comfortably | 114.49 tok/sEstimated | 2GB (have 12GB) |
| Qwen/Qwen3-4B-Instruct-2507 | Q8 | Fits comfortably | 82.11 tok/sEstimated | 4GB (have 12GB) |
| Qwen/Qwen3-4B-Instruct-2507 | FP16 | Fits comfortably | 43.50 tok/sEstimated | 9GB (have 12GB) |
| meta-llama/Llama-3.2-1B-Instruct | Q4 | Fits comfortably | 122.79 tok/sEstimated | 1GB (have 12GB) |
| meta-llama/Llama-3.2-1B-Instruct | Q8 | Fits comfortably | 93.43 tok/sEstimated | 1GB (have 12GB) |
| meta-llama/Llama-3.2-1B-Instruct | FP16 | Fits comfortably | 44.85 tok/sEstimated | 2GB (have 12GB) |
| openai/gpt-oss-120b | Q4 | Not supported | 20.74 tok/sEstimated | 59GB (have 12GB) |
| openai/gpt-oss-120b | Q8 | Not supported | 14.12 tok/sEstimated | 117GB (have 12GB) |
| openai/gpt-oss-120b | FP16 | Not supported | 8.45 tok/sEstimated | 235GB (have 12GB) |
| Qwen/Qwen2.5-3B-Instruct | Q4 | Fits comfortably | 133.21 tok/sEstimated | 2GB (have 12GB) |
| Qwen/Qwen2.5-3B-Instruct | Q8 | Fits comfortably | 88.09 tok/sEstimated | 3GB (have 12GB) |
| openai-community/gpt2 | Q4 | Fits comfortably | 102.45 tok/sEstimated | 4GB (have 12GB) |
Note: Performance estimates are calculated. Real results may vary. Methodology · Submit real data
Buy directly on Amazon with fast shipping and reliable customer service.
Rotate out primary variants whenever validation flags an issue.
Essential accessories to pair with RTX 5070
Total Bundle Price
All items from Amazon
💡 Not ready to buy? Try cloud GPUs first
Test RTX 5070 performance in the cloud before investing in hardware. Pay by the hour with no commitment.
Explore how RTX 4060 Ti 16GB stacks up for local inference workloads.
Explore how RX 6800 XT stacks up for local inference workloads.
Explore how RTX 4070 Super stacks up for local inference workloads.
Explore how RTX 3080 stacks up for local inference workloads.
Explore how RTX 3090 stacks up for local inference workloads.
RPG • 2020
RPG • 2023
Action RPG • 2023
RPG • 2023
Survival Horror • 2023
Action RPG • 2022
Action RPG • 2024
Action Adventure • 2025
Survival Horror • 2023
Action • 2022
Action Adventure • 2023
Action Adventure • 2019