RTX 4070 Ti: 12 GB, what it runs and what it costs
Specs checked · Rental prices from 30m ago · Aliteq Team · How we compute this
The short answer
The NVIDIA GeForce RTX 4070 Ti has 12 GB of VRAM and 504 GB/s of memory bandwidth (2023). At Q4_K_M quality or better with an 8K context it runs 14 of the open models we track; the largest that fits with room to spare is Gemma 3 12B. The cheapest rental right now is $0.092/hr (Vast.ai, interruptible).
- VRAM
- 12 GB
- Bandwidth
- 504 GB/s
- Board power
- 285 W
- Launched
- 2023
Source: NVIDIA specification · bandwidth not yet confirmed by a second source, so we don't estimate speed from it
Buy it or rent it
| Provider | Type | From, per hour | Rent |
|---|---|---|---|
| Vast.ai | Spot (can be interrupted) | $0.092/hr | Vast.ai |
| Vast.ai | On demand | $0.141/hr | Vast.ai |
| Runpod | On demand | $0.500/hr | Runpod |
Lowest price per provider in our hourly snapshot (30m ago). Rent buttons are referral links. Every card side by side: cloud GPU prices.
Which AI models run on the RTX 4070 Ti
Computed by our VRAM engine for an 8K context: the best quality each model fits at in 12 GB. Speeds are upper-bound estimates from memory bandwidth, not benchmarks.
| Model | Best fit | VRAM used | ~tok/s |
|---|---|---|---|
| DeepSeek R1 Distill 14B | Q5_0 | 12 GB98% · tight | — |
| Qwen2.5-Coder 14B | Q5_0 | 12 GB98% · tight | — |
| Qwen3 14B | Q5_K_M | 12 GB98% · tight | — |
| Phi-4 14B | Q5_0 | 12 GB98% · tight | — |
| Gemma 3 12B | Q6_K | 11 GB88% | — |
| Gemma 4 12B | Q6_K | 10 GB84% | — |
| Qwen3.5 9B | Q8_0 | 11 GB88% | — |
| Qwen3 8B | Q8_0 | 10 GB83% | — |
| Gemma 4 E4B | Q8_0 | 9 GB73% | — |
| Qwen2.5 7B Instruct | Q8_0 | 9 GB73% | — |
| Mistral 7B Instruct v0.3 | Q8_0 | 9 GB75% | — |
| Gemma 4 E2B | F16 | 10 GB86% | — |
Fits only below Q4 quality: Qwen3.5 27B (IQ3_XS), Qwen3.6 27B (IQ3_XS), Qwen3.8 27B (IQ3_XS), Gemma 4 26B-A4B (MoE) (Q2_K), Mistral Small 24B (Q2_K).
Too big for 12 GB (rent a bigger card instead): what it costs to run Qwen3 30B-A3B (MoE), what it costs to run Qwen3-Coder 30B-A3B (MoE), what it costs to run Gemma 4 31B, what it costs to run NVIDIA Nemotron 3 Nano 30B-A3B (MoE).
Check a model and context of your own in the VRAM calculator.
AI video on the RTX 4070 Ti
Alibaba's Wan 2.2 is an open video model you can run yourself in ComfyUI. With 12 GB the RTX 4070 Ti fits:
- Wan 2.2 5B, with ComfyUI's native offloading (8 GB and up, per ComfyUI's Wan 2.2 docs)
What each setup needs, the cheapest card to rent and what one clip costs: cloud GPUs for ComfyUI with Flux and Wan 2.2. Against an app or an API: AI video cost per clip, app vs API vs rented GPU. Rather not run a model yourself? Hosted video models such as Seedance, Veo and Hailuo are compared in Kling AI alternatives, priced per clip.
Cards to compare it with
- RTX 4070 SUPER · 12 GB
- RTX 4070 · 12 GB
- RTX 3060 · 12 GB
- RTX 5070 · 12 GB
- RTX 3080 Ti · 12 GB
- RTX 3080 · 12 GB
Every card by VRAM tier: all GPUs for gaming and local AI.