NVIDIA GeForce RTX 4070
12 GB · 504.2 GB/s · 200 W
In short
The NVIDIA GeForce RTX 4070 has 12 GB of GDDR6X memory with 504.2 GB/s of bandwidth, which sets how fast it generates text. At 8k context it runs 59 of the 125 open models we track fully in memory; the most capable is Qwen 3.8 27B at UD-Q6_K, at about 14 tok/s (estimated). We have no current price for the NVIDIA GeForce RTX 4070 in the US.
Updated . Speeds are estimates; how they are worked out.
Prices
Every retailer we track in the US, cheapest first. Buy links may earn us a commission.
No retailer in this market lists it right now.
What two of them run
Every model in the catalogue, with the best quant for this hardware. Estimates, not measurements.
| Model | Quant | Fits | Memory | Max context | Decode @8k | Prefill | Good for |
|---|---|---|---|---|---|---|---|
| Qwen 3.8 27B Alibaba Qwen | UD-Q6_K | fits | 22.3 GB | 15k27k with q8 cache | 14tok/sest. | 229 tok/s | – |
| Qwen 3.6 27B Alibaba Qwen | Q6_K | fits | 22.8 GB | 8k14k with q8 cache | 14tok/sest. | 229 tok/s | – |
| Gemma 4 31B | Q5_K_M | fits | 22.7 GB | 8k23k with q8 cache | 14tok/sest. | 200 tok/s | – |
| Qwen 3.6 35B-A3B Alibaba Qwen | Q4_K_M | fits | 22.1 GB | 34k54k with q8 cache | 90tok/sest. | 1966 tok/s | chat, coding, agents, long documents |
| Qwen 3.5 27B Alibaba Qwen | Q6_K | fits | 22.7 GB | 9k16k with q8 cache | 14tok/sest. | 229 tok/s | – |
| Gemma 4 26B-A4B | Q6_K | fits | 22.8 GB | 8k18k with q8 cache | 48tok/sest. | 1498 tok/s | chat, coding, agents, long documents |
| Qwen 3.5 35B-A3B Alibaba Qwen | Q4_K_M | fits | 21.9 GB | 44k70k with q8 cache | 91tok/sest. | 1966 tok/s | chat, coding, agents, long documents |
| GLM 4.7 Flash Z.ai | Q5_K_M | fits | 21.5 GB | 30k52k with q8 cache | 47tok/sest. | 1100 tok/s | chat, coding, agents |
| Devstral Small 2 24B Mistral AI | Q6_K | fits | 20.4 GB | 22k41k with q8 cache | 15tok/sest. | 259 tok/s | chat |
| gpt-oss 20B OpenAI | MXFP4 | fits | 12.6 GB | 128k | 71tok/sest. | 1423 tok/s | chat, coding, agents, long documents |
| Qwen3 Coder 30B-A3B Alibaba Qwen | Q5_K_M | fits | 22.1 GB | 14k26k with q8 cache | 64tok/sest. | 1422 tok/s | chat, coding |
| Qwen3 32B Alibaba Qwen | Q4_K_M | fits | 21.6 GB | 12k23k with q8 cache | 14tok/sest. | 177 tok/s | – |
| Gemma 4 12B | Q8_0 | fits | 13.4 GB | 256k | 24tok/sest. | 522 tok/s | chat, coding, long documents |
| Qwen 3.5 9B Alibaba Qwen | Q8_0 | fits | 10.3 GB | 256k | 32tok/sest. | 661 tok/s | chat, coding, long documents |
| Gemma 3 27B | Q5_K_M | fits | 20.1 GB | 39k74k with q8 cache | 15tok/sest. | 233 tok/s | chat |
| Mistral Small 3.2 24B Mistral AI | Q6_K | fits | 20.4 GB | 22k41k with q8 cache | 15tok/sest. | 259 tok/s | chat |
| DeepSeek R1 Distill Qwen 32B DeepSeek | Q4_K_M | fits | 21.6 GB | 12k23k with q8 cache | 14tok/sest. | 185 tok/s | – |
| Ministral 3 14B Mistral AI | Q8_0 | fits | 15.8 GB | 51k93k with q8 cache | 20tok/sest. | 429 tok/s | chat |
| Qwen3 14B Alibaba Qwen | Q8_0 | fits | 17.0 GB | 40k | 18tok/sest. | 399 tok/s | chat |
| DeepSeek R1 0528 Qwen3 8B DeepSeek | Q8_0 | fits | 10.4 GB | 92k128k with q8 cache | 32tok/sest. | 698 tok/s | chat, coding |
| Qwen3 8B Alibaba Qwen | Q8_0 | fits | 10.4 GB | 40k | 32tok/sest. | 698 tok/s | chat, coding |
| Phi-4 14B Microsoft | Q8_0 | fits | 17.2 GB | 16k | 18tok/sest. | 401 tok/s | chat |
| Qwen 3.5 4B Alibaba Qwen | Q8_0 | fits | 5.6 GB | 256k | 65tok/sest. | 1331 tok/s | chat, coding, long documents |
| Ministral 3 8B Mistral AI | Q8_0 | fits | 10.6 GB | 95k173k with q8 cache | 31tok/sest. | 652 tok/s | chat, coding |
| Granite 4.2 8B IBM | Q8_0 | fits | 11.1 GB | 80k128k with q8 cache | 29tok/sest. | 647 tok/s | chat, coding |
| Gemma 4 E4B | Q8_0 | fits | 9.0 GB | 128k | 37tok/sest. | 801 tok/s | chat, coding, long documents |
| LFM2.5 8B-A1B Liquid AI | Q8_0 | fits | 9.9 GB | 125k | 128tok/sest. | 4675 tok/s | chat, coding, long documents |
| Llama 3.1 8B Meta | Q6_K | fits | 8.3 GB | 118k128k with q8 cache | 41tok/sest. | 720 tok/s | chat, coding |
| Qwen2.5 0.5B Alibaba Qwen | Q8_0 | fits | 1.7 GB | 32k | 495tok/sest. | 9785 tok/s | chat, coding, long documents |
| Qwen3.5 0.8B Alibaba Qwen | Q8_0 | fits | 2.0 GB | 256k | 336tok/sest. | 6732 tok/s | chat, coding, agents, long documents |
| Qwen3.5 2B Alibaba Qwen | Q5_K_M | fits | 2.6 GB | 256k | 201tok/sest. | 2762 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 1.5B DeepSeek | Q8_0 | fits | 3.1 GB | 128k | 147tok/sest. | 3080 tok/s | chat, coding, long documents |
| MiniCPM5 2B openbmb | Q8_0 | fits | 4.0 GB | 128k | 103tok/sest. | 2036 tok/s | chat, coding, agents, long documents |
| LFM2.5 2.6B Liquid AI | Q8_0 | fits | 4.3 GB | 128k | 93tok/sest. | 2049 tok/s | chat, coding, agents, long documents |
| LFM2.5 VL 3B Liquid AI | Q8_0 | fits | 4.3 GB | 32k | 93tok/sest. | 1808 tok/s | chat, coding, agents, long documents |
| Spark X2.5 4B XHToken | mlx-4bit | fits | 4.3 GB | 458k767k with q8 cache | 92tok/sest. | 1449 tok/s | chat, coding, agents, long documents |
| LFM2.5 24B A2B Liquid AI | Q6_K | fits | 21.5 GB | 23k42k with q8 cache | 77tok/sest. | 2455 tok/s | chat, coding, agents |
| granite vision 4.1 4b IBM | Q8_0 | fits | 5.1 GB | 128k | 73tok/sest. | 1356 tok/s | chat, coding, long documents |
| Ministral 3 3B Instruct 2512 Mistral AI | Q8_0 | fits | 5.4 GB | 170k256k with q8 cache | 69tok/sest. | 1389 tok/s | chat, coding, agents, long documents |
| granite 4.2 3b IBM | Q8_0 | fits | 5.4 GB | 128k | 69tok/sest. | 1458 tok/s | chat, coding, agents, long documents |
| Mellum2.1 12B A2.5B Thinking JetBrains | Q8_0 | fits | 13.3 GB | 128k | 67tok/sest. | 2253 tok/s | chat, coding, agents, long documents |
| FrogNano 4B 2609 Microsoft | Q8_0 | fits | 5.7 GB | 256k | 63tok/sest. | 1331 tok/s | chat, coding, agents, long documents |
| Kimi VL A3B Thinking 2506 Moonshot AI | Q8_0 | fits | 17.2 GB | 128k | 56tok/sest. | 1899 tok/s | chat, coding, long documents |
| Kimi VL A3B Moonshot AI | mlx-8bit | fits | 18.0 GB | 128k | 53tok/sest. | 1899 tok/s | chat, coding, agents, long documents |
| DeepSeek R1 Distill Qwen 7B DeepSeek | Q8_0 | fits | 9.1 GB | 128k | 37tok/sest. | 777 tok/s | chat, coding, long documents |
| GLM 4 9B 0414 Z.ai | mlx-5bit | fits | 9.1 GB | 32k | 36tok/sest. | 611 tok/s | chat, coding |
| DeepSeek R1 Distill Llama 8B DeepSeek | Q8_0 | fits | 10.1 GB | 105k128k with q8 cache | 33tok/sest. | 720 tok/s | chat, coding |
| Ministral 3 8B Reasoning 2512 Mistral AI | Q8_0 | fits | 10.6 GB | 95k173k with q8 cache | 31tok/sest. | 652 tok/s | chat, coding |
| GLM 4.6V Flash Z.ai | Q8_0 | fits | 10.8 GB | 128k | 30tok/sest. | 564 tok/s | chat, coding |
| GLM 4.1V 9B Thinking Z.ai | mlx-8bit | fits | 12.4 GB | 64k | 26tok/sest. | 564 tok/s | chat, coding |
| Ministral 3 14B Reasoning 2512 Mistral AI | Q8_0 | fits | 15.8 GB | 51k93k with q8 cache | 20tok/sest. | 429 tok/s | chat |
| DeepSeek R1 Distill Qwen 14B DeepSeek | Q8_0 | fits | 17.3 GB | 36k67k with q8 cache | 18tok/sest. | 391 tok/s | chat |
| Devstral Small 2507 Mistral AI | Q5_K_M | fits | 18.0 GB | 37k68k with q8 cache | 17tok/sest. | 263 tok/s | chat |
| Magistral Small 2507 Mistral AI | Q5_K_M | fits | 18.0 GB | 37k40k with q8 cache | 17tok/sest. | 263 tok/s | chat |
| Magistral Small 2509 Mistral AI | Q5_K_M | fits | 18.0 GB | 37k68k with q8 cache | 17tok/sest. | 259 tok/s | chat |
| Muse Glimmer 30B meta-models | Q5_K_M | fits | 20.0 GB | 128k | 15tok/sest. | 213 tok/s | chat |
| Magistral Small 2506 Mistral AI | Q6_K | fits | 20.4 GB | 22k40k with q8 cache | 15tok/sest. | 263 tok/s | chat |
| Qwen3.8 27B OBLITERATED OBLITERATUS | mlx-4bit | fits | 21.0 GB | 33k59k with q8 cache | 15tok/sest. | 229 tok/s | – |
| granite 4.2 30b IBM | Q5_K_M | fits | 22.7 GB | 8k15k with q8 cache | 14tok/sest. | 209 tok/s | – |
| Qwen 3.5 122B-A10B Alibaba Qwen | UD-Q2_K_XL | offload | 40.4 GB | – | 16tok/sest. | 443 tok/s | chat |
| Qwen3 Coder Next Alibaba Qwen | Q4_K_M | offload | 46.6 GB | – | 20tok/sest. | 1051 tok/s | chat |
| Mistral Small 4 119B Mistral AI | UD-Q2_K_XL | offload | 38.8 GB | – | 24tok/sest. | 591 tok/s | chat |
| GLM 4.5 Air Z.ai | Q2_K | offload | 44.8 GB | – | 8tok/sest. | 252 tok/s | – |
| Qwen3 Next 80B-A3B Alibaba Qwen | Q4_K_M | offload | 46.6 GB | – | 26tok/sest. | 1322 tok/s | chat |
| Nemotron 3.5 Lightning 30B-A3B NVIDIA | MXFP4 | offload | 22.9 GB | 3k4k with q8 cache | 80tok/sest. | 2045 tok/s | chat, coding, agents, long documents |
| Kimi Linear 48B-A3B Moonshot AI | Q4_K_M | offload | 29.2 GB | – | 46tok/sest. | 1611 tok/s | chat, coding, long documents |
| Llama 3.3 70B Meta | Q4_K_M | offload | 43.2 GB | – | 2tok/sest. | 61 tok/s | – |
| DeepSeek R1 Distill Llama 70B DeepSeek | Q4_K_M | offload | 43.2 GB | – | 2tok/sest. | 61 tok/s | – |
| DeepSeek V4 Flash DeepSeek | UD-Q2_K_XL | won't run | 92.0 GB | – | – | – | – |
| GLM 5.3 Flash Z.ai | UD-Q2_K_XL | won't run | 102.6 GB | – | – | – | – |
| Qwen 3.8 Flash Next Alibaba Qwen | UD-Q2_K_XL | won't run | 74.9 GB | – | – | – | – |
| MiniMax M2.7 MiniMax | UD-Q2_K_XL | won't run | 73.2 GB | – | – | – | – |
| Nemotron 3 Super 120B-A12B NVIDIA | UD-Q2_K_XL | won't run | 52.3 GB | – | – | – | – |
| gpt-oss 120B OpenAI | Q4_K_M | won't run | 59.9 GB | – | – | – | – |
| DeepSeek R1 DeepSeek | Q4_K_M | won't run | 378.3 GB | – | – | – | – |
| DeepSeek R1 0528 DeepSeek | Q4_K_M | won't run | 378.8 GB | – | – | – | – |
| DeepSeek R1 Zero DeepSeek | Q4_K_M | won't run | 378.3 GB | – | – | – | – |
| DeepSeek V3 DeepSeek | Q4_K_M | won't run | 378.3 GB | – | – | – | – |
| DeepSeek V3 0324 DeepSeek | Q4_K_M | won't run | 378.3 GB | – | – | – | – |
| DeepSeek V3.1 DeepSeek | Q4_K_M | won't run | 379.2 GB | – | – | – | – |
| DeepSeek V3.1 Terminus DeepSeek | Q4_K_M | won't run | 379.2 GB | – | – | – | – |
| DeepSeek V3.2 DeepSeek | Q4_K_M | won't run | 379.2 GB | – | – | – | – |
| DeepSeek V4 Flash 0731 DeepSeek | MXFP4 | won't run | 147.5 GB | – | – | – | – |
| DeepSeek V4 Flash Vision Exp DeepSeek | MXFP4 | won't run | 156.2 GB | – | – | – | – |
| Devstral 2 123B Instruct 2512 Mistral AI | Q4_K_M | won't run | 73.7 GB | – | – | – | – |
| GLM 4.5 Z.ai | Q4_K_M | won't run | 205.6 GB | – | – | – | – |
| GLM 4.5V Z.ai | Q4_K_M | won't run | 61.8 GB | – | – | – | – |
| GLM 4.6 Z.ai | Q4_K_M | won't run | 204.8 GB | – | – | – | – |
| GLM 4.6V Z.ai | Q4_K_M | won't run | 68.2 GB | – | – | – | – |
| GLM 4.7 Z.ai | Q4_K_M | won't run | 205.6 GB | – | – | – | – |
| GLM 5 Z.ai | MXFP4 | won't run | 384.2 GB | – | – | – | – |
| GLM 5.1 Z.ai | MXFP4 | won't run | 421.8 GB | – | – | – | – |
| GLM 5.2 Z.ai | UD-Q4_K_M | won't run | 435.7 GB | – | – | – | – |
| GLM 5.3 Z.ai | Q8_0 | won't run | 748.2 GB | – | – | – | – |
| Kimi Dev 72B Moonshot AI | Q6_K | won't run | 63.6 GB | – | – | – | – |
| Kimi K2 Moonshot AI | Q4_K_M | won't run | 579.8 GB | – | – | – | – |
| Kimi K2 Instruct 0905 Moonshot AI | Q4_K_M | won't run | 580.3 GB | – | – | – | – |
| Kimi K2 Thinking Moonshot AI | Q4_K_M | won't run | 580.3 GB | – | – | – | – |
| Kimi K2.5 Moonshot AI | Q4_K_M | won't run | 580.3 GB | – | – | – | – |
| MiniMax M2 MiniMax | Q4_K_M | won't run | 131.9 GB | – | – | – | – |
| MiniMax M2.1 MiniMax | Q4_K_M | won't run | 131.9 GB | – | – | – | – |
| MiniMax M2.5 MiniMax | MXFP4 | won't run | 118.4 GB | – | – | – | – |
| MiniMax M3 MiniMax | MXFP4 | won't run | 241.8 GB | – | – | – | – |
| Mistral Medium 3.5 128B Mistral AI | Q4_K_M | won't run | 73.7 GB | – | – | – | – |
Specifications
From the manufacturer's published specifications; sources are linked where we have them.
- GPU
- GeForce RTX 4070
- Memory
- 12 GB GDDR6X
- Memory bus
- 192-bit
- Memory bandwidth
- 504.2 GB/s
- FP16 compute
- 29.146 TFLOPS
- Software
- CUDA
- Board power
- 200 W
- PCIe generation
- Gen 4
- PCIe lanes
- x16
Compare with
Questions
- How much memory does the NVIDIA GeForce RTX 4070 have for AI models?
- The NVIDIA GeForce RTX 4070 has 12 GB of GDDR6X memory with 504.2 GB/s of bandwidth, which sets how fast it generates text.
- What is the best model the NVIDIA GeForce RTX 4070 can run locally?
- Qwen 3.8 27B (UD-Q6_K) fits fully in its memory at 8k context and generates at about 14 tok/s, an estimate from its memory bandwidth. 59 of the 125 models we track fit at that context.
- Can the NVIDIA GeForce RTX 4070 run DeepSeek V4 Flash?
- No: even at UD-Q2_K_XL, DeepSeek V4 Flash needs about 92 GB at 8k context, more than the NVIDIA GeForce RTX 4070 and system RAM can hold.
- How much does the NVIDIA GeForce RTX 4070 cost in the US?
- We have no current price for the NVIDIA GeForce RTX 4070 in the US.
Speeds on this page are for 8k context. See the best hardware for local AI at every budget.