UNIT · NVIDIA · GPU
11 GB VRAMhighReviewed May 2026

NVIDIA GeForce GTX 1080 Ti

Pascal halo card. 11 GB GDDR5X at 484 GB/s — outperforms many newer mid-range cards on raw bandwidth. Runs 7B Q4 at ~50-65 tok/s, 13B Q4 fits comfortably at ~25-35 tok/s. The legendary 'still relevant' card for AI on a budget; used $230-280 makes it the value flagship of 2026.

Released 2017·~$250 street·484 GB/s memory bandwidth
RUNLOCALAI SCORE
See full leaderboard →
327/ 1000
CC-tier
Estimated
Throughput
168/ 500
VRAM-fit
80/ 200
Ecosystem
200/ 200
Efficiency
19/ 100

Extrapolated from 484 GB/s bandwidth — 58.1 tok/s estimated. No measured benchmarks yet.

Plain-English: Best for 7B; 14B is tight — coding agent feels deliberate; vision models supported.

7B chat
Comfortable
14B chat~
Tight
32B chat
Doesn't fit
70B chat
Doesn't fit
Coding agent~
Tight
Vision (≤8B VLM)
Comfortable
Long context (32K)~
Tight
Comfortable — fits with headroom
~Tight — works, no slack
Marginal — needs aggressive quant
Doesn't fit usefully

Verdicts extrapolated from catalog VRAM + bandwidth + ecosystem flags. Hover any chip for the rationale. Want measured numbers? Submit your own run with runlocalai-bench --submit.

BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 10, 2026
6.6/10

This card is for the operator who needs a capable local inference rig on a strict budget, and is willing to accept older architecture limitations. The GTX 1080 Ti runs 7B Q4 models at ~50-65 tok/s and 13B Q4 at ~25-35 tok/s, making it a strong performer for chat and code completion workloads. Its 11 GB VRAM fits 13B Q4 comfortably, and even 30B Q3 models can be squeezed in with careful quantization. However, larger models like 34B or 70B are out of reach, and the lack of FP16 tensor cores means slower performance on mixed-precision inference. CUDA support is solid, but newer features like Flash Attention may not be fully optimized. Pass on this card if you need to run 70B+ models, require FP16 throughput for training, or want the latest software compatibility. At $250 used, it's the value champion for 7B-13B inference, but expect to upgrade when larger models become your daily driver.

Why this rating

The GTX 1080 Ti offers exceptional price-to-performance for 7B-13B inference, with high bandwidth and sufficient VRAM for its era. It loses points for lack of modern features and inability to handle larger models, but remains a top budget pick.

BLK · OVERVIEW

Overview

Pascal halo card. 11 GB GDDR5X at 484 GB/s — outperforms many newer mid-range cards on raw bandwidth. Runs 7B Q4 at ~50-65 tok/s, 13B Q4 fits comfortably at ~25-35 tok/s. The legendary 'still relevant' card for AI on a budget; used $230-280 makes it the value flagship of 2026.

Retailers we'd check:Amazon

Search-fallback links. Editorial hasn't yet curated retailer URLs for this card. Approx. $250.

Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.

BLK · SPECS

Specs

VRAM11 GB
Power draw250 W
Released2017
MSRP$699
Backends
CUDA
Vulkan

Models that fit

Open-weight models small enough to run on NVIDIA GeForce GTX 1080 Ti with usable context.

Compare alternatives

Hardware worth comparing

Same VRAM tier and the one step above and below — so you can frame the buying decision against real options.

Frequently asked

What models can NVIDIA GeForce GTX 1080 Ti run?

With 11GB VRAM, the NVIDIA GeForce GTX 1080 Ti runs models up to 14B in 4-bit, or 7B at higher quantizations. See the model list below for tested combinations.

Does NVIDIA GeForce GTX 1080 Ti support CUDA?

Yes — NVIDIA GeForce GTX 1080 Ti is an NVIDIA card with full CUDA support, the most mature local-AI backend. llama.cpp, Ollama, vLLM, and ExLlamaV2 all run natively.

How much does NVIDIA GeForce GTX 1080 Ti cost?

Current street price for NVIDIA GeForce GTX 1080 Ti is around $250 (MSRP $699). Prices vary by region and supply.

Where next?

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify hardware specifications.