GeForce RTX 5070 Ti Price & Used Prices: Which Local LLMs Can It Run?
Price alert
We'll email you when the eBay (US) price reaches your target.
The RTX 5070 Ti is NVIDIA's 2025 16GB gaming card: 12B-class at 4-bit runs about 83 tokens/s, but 27B won't fit. No Founders Edition; measure your case.
eBay
GeForce RTX 5070 Ti price history
eBay
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Full GeForce RTX 5070 Ti specifications
- Memory16 GB · GDDR7
- Bandwidth896 GB/s
- FP16 / BF16175.8 TFLOPS
- FP8 / INT8351.5 / 351.5 TOPS
- Bus width
- 256-bit
- Architecture
- Blackwell (GB203)
- Cores
- 8,960
- Tensor cores
- 280
- Ecosystem
- CUDA
- Interface
- PCIe 5.0 x16
- NVLink
- Not supported
- Power
- 300 W
- Power connector
- 16-pin 12V-2x6
- Type
- Consumer
- Launch date
- Feb 20, 2025
- MSRP
- $749
Best GeForce RTX 5070 Ti deals on eBay
Hand-picked deals from trusted eBay sellers. Shop with confidence.
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
UsedGigabyte RTX 5070 Ti WindForce SFF 16G$1,050+ shippingBuy on eBay
UsedLenovo NVIDIA GeForce RTX 5070 Ti 16GB GDDR7 Dedicated Graphics Card GPU - UsedTop Rated$1,110$1,725-36%Buy on eBay
UsedASUS TUF RTX 5070 Ti White OC Edition 16GB GDDR7 Graphics Card$1,150+ shippingBuy on eBay
UsedGIGABYTE GeForce RTX 5070 Ti Eagle OC SFF 16G Graphics Card NVIDIA$1,180+$18 shippingBuy on eBay
UsedASUS TUF Gaming GeForce RTX 5070 Ti OC 16GB GDDR7 / Gaming Graphics Card$1,200Buy on eBay
UsedMSI GEFORCE RTX 5070 TI VENTUS 3X OC 16GB GDDR7 Nvidia Graphics Card 5070tiTop Rated$1,249Buy on eBay
What LLMs can a single GeForce RTX 5070 Ti run?
A single 5070 Ti runs up to 12B at 4-bit or 8-bit. Bigger needs 2-bit: Qwen3.6 35B A3B runs about 158 tokens/s, triple the speed of using system RAM.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 E4B8B | 3.8 GB | Runs | 142 | 128K |
 Qwen3.5 9B9.7B | 4.1 GB | Runs | 132.3 | 256K |
| Gemma 4 12B12B | 4.7 GB | Runs | 116.6 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 137.5 | 256K |
| Qwen3.6 27B27.8B | 11.8 GB | Runs | 52.7 | 64K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 62.3 | 93K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 148.2 | 75K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 50.9 | 41K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 158.4 | 181K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 5 GB | Runs | 113.8 | 128K |
| Qwen3.5 9B9.7B | 5.7 GB | Runs | 101.1 | 256K |
| Gemma 4 12B12B | 7.1 GB | Runs | 82.8 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Needs RAM offload | 38.6 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 16.5 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Needs RAM offload | 36.5 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Needs RAM offload | 51.9 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 8.2 GB | Runs | 74.3 | 128K |
| Qwen3.5 9B9.7B | 9.5 GB | Runs | 64.8 | 196K |
| Gemma 4 12B12B | 12.7 GB | Runs | 49.4 | 183K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Needs RAM offload | 26.1 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 29 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Needs RAM offload | 22.8 | 198K |
| Gemma 4 31B30.7B | 32.6 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Needs RAM offload | 34.6 | 256K |
RunsFits fully in VRAM; speed is estimated from VRAM bandwidth.
Needs RAM offloadMoE only: expert weights sit in system RAM, so speed is estimated from 70 GB/s RAM bandwidth.
Too bigWon’t fit on one card at this quantization, even with RAM offload.
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- Mainly running models up to 12B
Gemma 4 12B runs at about 83 tokens/s at 4-bit and about 49 at 8-bit, both with more than 180K of context
- Running MoE models at 2-bit entirely on the card
Qwen3.6 35B A3B runs at about 158 tokens/s, three times as fast as at 4-bit with its experts in system RAM
- You want a new card with a warranty and speed close to a 3090
On dense models both cards can hold, it is only 4% slower, and it draws 50W less
Not a good fit
- Running 27B to 31B dense models at 4-bit
Qwen3.8 27B needs 16.5 GB and Gemma 4 31B needs 18.3 GB, and neither fits
- Running 30B-class MoE models at 4-bit
Qwen3.6 35B A3B has to put its experts in system RAM and runs at about 52 tokens/s, while a 3090, which holds it entirely, does about 135
Before you buy
No Founders Edition, so size depends on the model
NVIDIA made no Founders Edition of the 5070 Ti. Every card comes from a board partner, and length, thickness and cooling vary by brand, so measure your case against the specific model before you build. For the power supply, follow NVIDIA's recommendation of 750W.
Check the ROP count in GPU-Z
In February 2025 NVIDIA confirmed that fewer than 0.5% of 50-series cards shipped missing some of their ROPs (render output units), making them about 4% slower in games on average. NVIDIA says AI compute is not affected, and affected cards can be replaced through the manufacturer. A normal 5070 Ti has 96 ROPs; when buying used, ask the seller for a GPU-Z screenshot.
What to do when 16GB isn't enough
First shorten the context, then drop to 3-bit or 2-bit; with MoE models you can also put the experts in system RAM. Only if none of that is enough should you consider a 24GB card.
Compared with similar GPUs
GeForce RTX 5070 Ti vs GeForce RTX 3090, GeForce RTX 5080, and Radeon RX 9070 XT: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | GeForce RTX 5070 Ti | GeForce RTX 3090 | GeForce RTX 5080 | Radeon RX 9070 XT |
|---|---|---|---|---|
| Memory | 16 GB | 24 GB50% more | 16 GBSame | 16 GBSame |
| Bandwidth | 896 GB/s | 936 GB/s4% higher | 960 GB/s7% higher | 640 GB/s29% lower |
| BF16 / FP16 compute | 176 TFLOPS | 142 TFLOPS19% lower | 225 TFLOPS28% higher | 195 TFLOPS11% higher |
| INT8 compute | 352 TOPS | 285 TOPS19% lower | 450 TOPS28% higher | 389 TOPS11% higher |
| Qwen3.8 27B speed (2-bit) | 62.3 t/s | 64.8 t/s4% faster | 66.3 t/s6% faster | 45.6 t/s27% slower |
| Largest dense model (4-bit) | 23B | 37B61% more | 23BSame | 23BSame |
| Power | 300 W | 350 W17% higher | 360 W20% higher | 304 WSame |
| GPU interconnect | Not supported | NVLink 112.5 GB/s | Not supported | — |
| Interface | PCIe 5.0 x16 | PCIe 4.0 x16 | PCIe 5.0 x16 | PCIe 5.0 x16 |
| Released | 2025-02 | 2020-09 | 2025-01 | 2025-03 |
| eBay median |
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
GeForce RTX 5070 Ti rental price per hour
On Vast.ai, the GeForce RTX 5070 Ti rents for $0.54 per hour on demand, about $131/month at 8 hours a day. Buying one (eBay median $1,365) pays off in about 1.9 years, 1 year or 4 months at 4, 8 or 24 hours a day.
| Option | Price | 4 h/day | 8 h/day | 24 h/day | View |
|---|---|---|---|---|---|
 Vast.aiRentUpdated 48 min. ago | $0.54/hr | $65/mo | $131/mo | $393/mo | Rent |
 eBayUpdated 6 hr. ago | $1,365 | Payback ~1.9 yrsPower $7/mo | Payback ~1 yrPower $13/mo | Payback ~4 moPower $39/mo | Buy |
Break-even uses the US residential average ($0.18/kWh) and 300 W rated power, excluding resale value and the rest of the build.
Rental links are referral links: we may receive a commission or credit if you sign up through them. As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Amazon.jp
Xianyu