NVIDIA RTX 5070 Ti
The NVIDIA RTX 5070 Ti has 16 GB VRAM and 896 GB/s memory bandwidth. It can run 38 of our 84 tracked models natively in VRAM at 8k context.
With 16 GB GDDR7, the NVIDIA RTX 5070 Ti is a consumer-tier GPU that can run 38 models natively. It handles smaller models (7B–14B) at Q4–Q5 quantization.
The NVIDIA RTX 5070 Ti shares the same 16GB GDDR7 capacity as the 5080 but on slightly slower 28 Gbps memory (896 GB/s). With 8,960 CUDA cores and a 300W TDP, it hits a compelling price-to-performance point for 1440p gaming and can run 7B–14B LLMs entirely in VRAM.
NVIDIA RTX 5070 Ti: February 2025 Blackwell GB203 die with 16GB GDDR7 at 896 GB/s — $749 MSRP, same VRAM as the 5080 on slightly slower memory.
7B-14B models fit natively at Q4-Q6, matching the 5080's model compatibility with roughly 10% lower throughput. ~10-16 t/s for 7B Q4.
Full CUDA support. The best price-to-VRAM ratio in the Blackwell consumer lineup for 16GB-class local LLM work.
| Vendor | NVIDIA |
| Architecture | Blackwell |
| VRAM | 16 GB |
| Memory type | GDDR7 |
| Memory bandwidth | 896 GB/s |
| Compute backend | CUDA |
| Tier | Consumer |
| Released | 2025 |
| Models (native) | 38 / 84 |
| Models (offload) | 13 / 84 |
Software: Full llama.cpp and Ollama support out of the box. CUDA 12.x recommended; driver ≥ 525 required.
Popular models for this GPU
Models this GPU runs natively in VRAM (38)
- Qwen 3.5 35B-A3B (MoE)35B · MMLU-Pro 85.3Q2_K · ~146.4 t/s
- Nemotron 3 Nano 30B32B · MMLU-Pro 78.3Q2_K · ~137.2 t/s
- Qwen3 30B-A3B (MoE)30B · MMLU-Pro 61.5Q2_K · ~126.2 t/s
- Gemma 2 27B Instruct27.2B · MMLU-Pro 38.0Q2_K · ~43.3 t/s
- Gemma 3 27B Instruct27B · MMLU-Pro 67.5Q2_K · ~49.2 t/s
Show 33 more
- Qwen 3.6 27B27B · MMLU-Pro 86.2Q3_K_M · ~43.1 t/s
- Bonsai 27B27B · MMLU-Pro 81.5Ternary (Q2_0) · ~66.1 t/s
- Gemma 4 26B (MoE)26B · MMLU-Pro 82.6NVFP4 · ~86.8 t/s
- Mistral Small 3.1 24B Instruct24B · MMLU-Pro 66.8NVFP4 · ~43.7 t/s
- Mistral Small 22B22.2B · MMLU-Pro 49.2NVFP4 · ~44.9 t/s
- GPT-OSS 20B21B · MMLU-Pro 67.9NVFP4 · ~93.9 t/s
- Qwen3 14B14.8B · MMLU-Pro 61.0NVFP4 · ~66.6 t/s
- Qwen 2.5 14B Instruct14.7B · MMLU-Pro 63.7NVFP4 · ~65 t/s
- Phi-4 14B Instruct14B · MMLU-Pro 70.4NVFP4 · ~69.8 t/s
- Mistral Nemo 12B Instruct12.2B · MMLU-Pro 35.6NVFP4 · ~78.3 t/s
- Gemma 3 12B Instruct12.2B · MMLU-Pro 60.6NVFP4 · ~81.4 t/s
- Gemma 4 12B (Unified)12B · MMLU-Pro 77.2NVFP4 · ~63.2 t/s
- Gemma 2 9B Instruct9.2B · MMLU-Pro 32.0NVFP4 · ~78.5 t/s
- Qwen 3.5 9B9B · MMLU-Pro 82.5NVFP4 · ~122.1 t/s
- Llama 3.1 8B Instruct8B · MMLU-Pro 48.3NVFP4 · ~114.8 t/s
- DeepSeek R1 Distill Llama 8B8B · MMLU-Pro 41.0NVFP4 · ~114.8 t/s
- Qwen3 8B8B · MMLU-Pro 56.7NVFP4 · ~111.8 t/s
- Qwen 2.5 7B Instruct7.6B · MMLU-Pro 56.3NVFP4 · ~136.4 t/s
- Mistral 7B Instruct v0.37.25B · MMLU-Pro 30.0NVFP4 · ~123.9 t/s
- Gemma 3 4B Instruct4B · MMLU-Pro 43.6BF16 · ~68.5 t/s
- Gemma 4 E4B4B · MMLU-Pro 69.4BF16 · ~64.7 t/s
- Phi-3.5 Mini Instruct3.8B · MMLU-Pro 47.4BF16 · ~53.8 t/s
- Phi-4-mini Instruct3.8B · MMLU-Pro 67.3BF16 · ~67.1 t/s
- Llama 3.2 3B Instruct3.2B · MMLU-Pro 24.0BF16 · ~79.4 t/s
- Qwen 2.5 3B Instruct3.1B · MMLU-Pro 32.4FP32 · ~45.9 t/s
- Gemma 2 2B Instruct2.6B · MMLU-Pro 17.8FP32 · ~51.7 t/s
- Gemma 4 E2B2B · MMLU-Pro 60.0FP32 · ~69.3 t/s
- SmolLM2 1.7B Instruct1.7B · MMLU-Pro 19.0FP32 · ~69.2 t/s
- Qwen 2.5 1.5B Instruct1.5B · MMLU-Pro 16.8FP32 · ~93.4 t/s
- Llama 3.2 1B Instruct1.24B · MMLU-Pro 12.5FP32 · ~111.4 t/s
- Gemma 3 1B Instruct1B · MMLU-Pro 14.7FP32 · ~134.6 t/s
- Qwen 2.5 0.5B Instruct0.5B · MMLU-Pro 10.0FP32 · ~277.2 t/s
- SmolLM2 360M Instruct0.36B · MMLU-Pro 8.0FP32 · ~328 t/s
Models that fit with CPU offload (13)
These use system RAM for layers that don't fit in VRAM — expect much slower inference.
- Qwen 2.5 72B Instruct72B · MMLU-Pro 71.1Q2_K · ~1.6 t/s
- Llama 3.3 70B Instruct70B · MMLU-Pro 68.9Q3_K_M · ~1.1 t/s
- DeepSeek R1 Distill Llama 70B70B · MMLU-Pro 70.0Q3_K_M · ~1.1 t/s
- Llama 3.1 70B Instruct70B · MMLU-Pro 66.4Q3_K_M · ~1.1 t/s
- Mixtral 8x7B Instruct v0.146.7B · MMLU-Pro 29.7NVFP4 · ~2.7 t/s
- Command-R 35B35B · MMLU-Pro 33.0NVFP4 · ~1.8 t/s
- Qwen 3.6 35B35B · MMLU-Pro 85.2NVFP4 · ~4.9 t/s
- Yi 1.5 34B Chat34.4B · MMLU-Pro 37.0NVFP4 · ~5.4 t/s
- Qwen3 32B32.8B · MMLU-Pro 65.5NVFP4 · ~7.8 t/s
- Qwen 2.5 32B Instruct32.5B · MMLU-Pro 69.0NVFP4 · ~6.5 t/s
- Qwen 2.5 Coder 32B Instruct32.5B · MMLU-Pro 50.4NVFP4 · ~6.5 t/s
- DeepSeek R1 Distill Qwen 32B32.5B · MMLU-Pro 65.0NVFP4 · ~6.5 t/s
- Gemma 4 31B31B · MMLU-Pro 85.2NVFP4 · ~6 t/s
Too large for this GPU (33)
- Mixtral 8x22B Instruct v0.1
- Llama 3.1 405B Instruct
- DeepSeek V3 671B
- DeepSeek R1 671B
- Llama 4 Scout 109B
- Llama 4 Maverick 400B
- Qwen3 235B-A22B (MoE)
- MiniMax M1 456B
- GPT-OSS 120B
- GLM-4.5 355B
- GLM-4.5 Air 106B
- GLM-4.6 355B
- GLM-4.6V 106B
- GLM-4.7 358B
- Qwen 3.5 122B-A10B (MoE)
- MiniMax M2.5 229B
- GLM-5 744B
- MiniMax M2.7 229B
- Nemotron 3 Super 120B
- Kimi K2.6
- GLM-5.1 754B
- DeepSeek V4 Pro 1.6T
- DeepSeek V4 Flash 284B
- Mistral Medium 3.5 128B
- GLM-5.2 753B
- Nemotron 3 Ultra 550B-A55B
- Step 3.5 Flash
- Step 3.7 Flash
- MiMo V2.5 Pro
- Kimi K2.5
- MiniMax M3
- Inkling
- Kimi K3
Frequently asked questions
- How much VRAM does the NVIDIA RTX 5070 Ti have?
- The NVIDIA RTX 5070 Ti has 16 GB of GDDR7 with 896 GB/s memory bandwidth.
- What is the NVIDIA RTX 5070 Ti best for?
- With 16 GB of VRAM, the NVIDIA RTX 5070 Ti handles smaller models (7B–14B) at Q4–Q5 quantization — ideal for entry-level local LLM experimentation and lightweight inference.
- What LLMs can the NVIDIA RTX 5070 Ti run locally?
- The NVIDIA RTX 5070 Ti can run 38 of the 84 open-weight models tracked by CanItRun natively in VRAM at 8k context. Top options include: Llama 3.1 8B Instruct at NVFP4, Llama 3.2 3B Instruct at BF16, Llama 3.2 1B Instruct at FP32.
- Can the NVIDIA RTX 5070 Ti run Llama 3.3 70B Instruct?
- The NVIDIA RTX 5070 Ti can run Llama 3.3 70B Instruct with CPU offload at Q3_K_M quantization, but inference will be slower than native VRAM execution.
- Can the NVIDIA RTX 5070 Ti run Qwen 3.6 27B?
- Yes. The NVIDIA RTX 5070 Ti runs Qwen 3.6 27B natively in VRAM at Q3_K_M quantization, achieving approximately 43.1 tokens per second.
- Can the NVIDIA RTX 5070 Ti run Llama 3.1 8B Instruct?
- Yes. The NVIDIA RTX 5070 Ti runs Llama 3.1 8B Instruct natively in VRAM at NVFP4 quantization, achieving approximately 114.8 tokens per second.