RTX 5070 Ti
An RTX 5070 Ti runs 12 of the 34 open models tracked here at Q4_K_M with 8k context, computed on 16 October 2026. The largest is gpt-oss 20B, needing 13.4 GB and generating at most 300 tokens per second.
- Memory
- 16 GB VRAM
- Bandwidth
- 896 GB/s
- Board power
- 300 W
- Released
- 2025
- Availability
- Sold new
Open the RTX 5070 Ti in the LLM checker
What an RTX 5070 Ti runs, model by model
RTX 5070 Ti (16 GB) addresses 16 GB at 896 GB/s. Every model below is at Q4_K_M with 8k of context, the setting most people actually use. Once the 5% safety margin comes off, the card has 15.2 GB to spend: memory decides what loads, bandwidth how fast it answers. Both are estimates from stated formulas, and the VRAM calculator shows the working.
On this card a model fits at or under 15.2 GB, and above 13.6 GB it is tight: it loads, but there is little left for the conversation, and a desktop on the same 16 GB can push it over. Tokens per second are single-stream estimates from memory bandwidth; ≤ marks a ceiling real runtimes fall well below.
Computed 16 October 2026The model to actually run on it
Gemma 4 12B is the best use of this card: 8.2 GB of the 16 GB available, an estimated 90 tokens per second, faster than you can read, and 131,072 tokens of context still available.
How this is chosen, since no benchmark is quoted: models are compared by size, a mixture-of-experts model counting at the geometric mean of its total and active parameters (a rule of thumb, not a measurement). The pick is the biggest class that fits with memory to spare and answers at 25 tokens per second or better, and within that class the most recent general-purpose release; coding and reasoning specialists are listed above but not recommended by default.
gpt-oss 20B is the largest model the card will hold, at 13.4 GB, with room for 81,920 tokens of context. It is a mixture-of-experts model reading only 3.6B parameters per token, so it is the faster of the two at an estimated 300 tokens per second; what it gives up is the depth of a dense model that reads all of its weights for every token.
It also fits at Q8_0, in about 14.2 GB, which is worth taking whenever the memory allows: Q8 is near-lossless where Q4 costs a little accuracy.
How much context actually fits
Model size is the question people ask; context is the one that bites. The KV cache grows linearly with the conversation, so a model that loads comfortably can still run out of memory halfway through a long document. On 16 GB, the context you get is whatever the weights leave behind, which is why two cards that run the same model can feel completely different in use. These are the largest contexts this card holds, at Q4_K_M weights.
| Model | Max context, FP16 KV | With Q8 KV cache |
|---|---|---|
| Gemma 4 12B | 131,072 tokens | 131,072 tokens |
| Qwen3 14B | 34,816 tokens | 40,960 tokens · model limit |
| Phi-4 14B | 16,384 tokens · model limit | 16,384 tokens · model limit |
| Ministral 3 14B | 37,888 tokens | 75,776 tokens |
Quantising the KV cache to Q8 roughly doubles what you can hold, at a quality cost most people never notice: Ministral 3 14B on this card goes from 37,888 to 75,776 tokens. Capped at 128k tokens, or at the model's own native window where that is smaller, marked "model limit": free memory beyond that point buys nothing. Rows that hold far more context than their size suggests are hybrid, sliding-window or latent-attention models, which cache only a fraction of what a standard transformer does; each model page shows the working.
Where this card stops
The first model out of reach is Devstral Small 2 24B: about 16.3 GB at Q4_K_M and 8k context, against 16 GB of usable memory. Dropping to Q3_K_M would need about 13.6 GB, which fits, though Q3 loses enough quality that a smaller model at Q4 is usually the better trade. Offloading the remainder to system RAM works and is ten to fifty times slower; it is a way to see a model run, not a way to use one.
Buy or rent
The sensible 16 GB card if you want new and fast: it gives up little to the 5080 on language models. If you need more than 16 GB, a used 3090 holds 24 GB at about the same speed ceiling, at higher power. For the occasional larger model, renting by the hour is cheaper than buying up a tier.
Retail prices drift, and a figure written today would be wrong within a month, so this page does not carry one. Put the figure you are actually looking at, with your electricity rate and your real monthly hours, into the build vs rent calculator, which already knows this card draws 300 W under load. It returns the month owning becomes cheaper, or tells you it never does. Hours per month decides it far more often than the price does.
The honest take
The 5080's memory bus and 16 GB, read about 7% more slowly.
It has the same 16 GB on the same 256-bit bus as the 5080 and reads it at 896 GB/s against 960 GB/s, so on language models the two cards are closer than their names suggest. That also puts it within about 5% of a used 3090, which holds 24 GB. It is a Blackwell card: FP4 formats run natively, and old software builds will not use it.
What it is good at
- 896 GB/s: twice the 5060 Ti 16 GB's 448 GB/s, at the same capacity.
- Within about 7% of the 5080 on anything that fits, because the memory bus is the same width.
- Blackwell generation, so FP4 weight formats run natively.
What it is not
- Sixteen gigabytes: it holds what any 16 GB card holds, and less than a used 3090.
- Blackwell needs CUDA 12.8 or newer builds of PyTorch and llama.cpp; an old install will not use the card.
- NVIDIA makes no Founders Edition of this card, so size and cooling vary by maker; board power is 300 W.
The thing people get wrong: The 5070 Ti and the 5080 share the 256-bit bus and the 16 GB, and the 5080 reads memory only about 7% faster. For chat with a model that fits, the gap is small. Most of what the 5080 adds is compute, which shows in image generation and prompt processing rather than in generation speed.
Compared with the alternatives
- RTX 5080: 16 GB · 960 GB/s · best fit Gemma 4 12B at ~97 tok/s. A 16 GB card with 3090-class bandwidth: quick on what fits, and it fits no more than a 5060 Ti 16 GB.
- RTX 3090: 24 GB · 936 GB/s · best fit Muse Glimmer 30B at ~38 tok/s. The value benchmark for local models, and it has been for years.
- RTX 5060 Ti: 16 GB · 448 GB/s · best fit Gemma 4 12B at ~45 tok/s. The newest 16 GB GeForce below the high end, and a clear step up from the 4060 Ti it replaces.
All 39 cards are compared side by side on the GPU index, and the same numbers from the model's point of view are on each model fit table. To check one specific pairing rather than read a table, use the can-I-run-it checker.
Questions
- Can an RTX 5070 Ti run a 70B model?
- Not at Q4_K_M in one card. Llama 3.3 70B needs about 45.8 GB at 8k context and this card can address 16 GB. Your options are a smaller model, a harsher quantisation with very little context, splitting across two cards, or renting a larger one for the hours you need it.
- What is the best model to run on an RTX 5070 Ti?
- Gemma 4 12B. At Q4_K_M and 8k context it needs about 8.2 GB of the 16 GB available and generates an estimated 90 tokens per second, which is faster than you can read. It also fits at Q8, at about 14.2 GB, which is worth taking when it fits.
- How many tokens per second does an RTX 5070 Ti generate?
- It depends entirely on the size of the model, because generating a token means reading every active weight out of memory. At 896 GB/s and 70% efficiency, this card produces an estimated 135 tokens per second on an 8B model at Q4, and about 300 on the largest model it holds, gpt-oss 20B. Reading speed is roughly 5 to 8 words per second, so anything above 25 feels immediate. Treat 135 as a ceiling on 896 GB/s rather than a measurement; the higher a figure, the further real runtimes fall below it.
- Is 16 GB enough for running LLMs locally?
- It runs 12 of the 34 models tracked here at Q4_K_M with 8k of context, up to 21B parameters. The honest test is not the model list but the context: Gemma 4 12B on this card holds about 131,072 tokens before memory runs out.
- RTX 5070 Ti or RTX 5080 for local models?
- Both address about the same memory, so they run the same models. On speed, the RTX 5080 is faster: 960 GB/s against 896 GB/s, and bandwidth is what sets chat speed.
- RTX 5070 Ti or RTX 3090 for AI?
- The 3090 holds more, 24 GB against 16 GB, and reads it slightly faster, 936 GB/s against 896 GB/s, so it fits larger models and generates about as fast on anything both hold. The 5070 Ti is new, draws 300 W against 350 W, and runs FP4 formats natively. If capacity decides it, the 3090; if you want a new card with current features, the 5070 Ti.
- RTX 5070 Ti or RTX 5080 for AI?
- Both have 16 GB on a 256-bit bus, so the same models fit. The 5080 reads memory at 960 GB/s against 896 GB/s, about 7% faster, and for chat that is the whole difference. Its extra compute helps image and video generation and long prompts. For language models alone, the 5070 Ti gives up very little.
- Memory: weights (parameters × bytes per parameter) + KV cache (2 × layers × KV heads × head dim × context × bytes for a standard transformer; sliding-window, hybrid and latent-attention models counted as they actually cache) + 0.5 GB runtime + 4% of weights. Full derivation in the VRAM calculator.
- Speed: 70% × bandwidth ÷ active weight bytes, single stream, no batching, prompt processing excluded. A ceiling, not a measurement: the 7 of 12 figures marked ≤ are where real runtimes fall furthest below the number, because so few weights are read per token. See the tokens-per-second estimator.
- Card specification (manufacturer figures): 16 GB, 896 GB/s, 300 W; see NVIDIA's own page for this card. The bandwidth figure: NVIDIA's GeForce comparison page.
- Model architecture values come from each model's published config; the self-hosted LLM guide explains what each one changes.