A100
An A100 runs 34 of the 34 open models tracked here at Q4_K_M with 8k context, computed on 16 October 2026. The largest is Mistral Small 4 119B (MoE), needing 72.5 GB and generating at most 379 tokens per second.
- Memory
- 80 GB VRAM
- Bandwidth
- 2,039 GB/s
- Released
- 2020
- Availability
- Rented, not bought
Open the A100 in the LLM checker
What an A100 runs, model by model
A100 80 GB addresses 80 GB at 2,039 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 76 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 76 GB, and above 68 GB it is tight: it loads, but there is little left for the conversation, and a desktop on the same 80 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
Llama 3.3 70B is the best use of this card: 45.8 GB of the 80 GB available, an estimated 35 tokens per second, comfortable for chat, and 100,352 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.
Mistral Small 4 119B (MoE) is the largest model the card will hold, at 72.5 GB, with room for 131,072 tokens of context even at 91% of memory, because its attention design caches so little per token. It is a mixture-of-experts model reading only 6.5B parameters per token, so it is the faster of the two at an estimated 379 tokens per second; what it gives up is the depth of a dense model that reads all of its weights for every token.
If quality matters more than parameter count, Muse Glimmer 30B fits at Q8_0 in about 33.5 GB. Q8 is near-lossless; a smaller model at Q8 often beats a larger one squeezed into Q3.
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 80 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 |
|---|---|---|
| Llama 3.3 70B | 100,352 tokens | 131,072 tokens |
| Qwen3 32B | 40,960 tokens · model limit | 40,960 tokens · model limit |
| Gemma 4 31B | 131,072 tokens | 131,072 tokens |
| Muse Glimmer 30B | 131,072 tokens | 131,072 tokens |
Quantising the KV cache to Q8 roughly doubles what you can hold, at a quality cost most people never notice: Llama 3.3 70B on this card goes from 100,352 to 131,072 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
Nothing in this list is out of reach. Every model tracked here fits at Q4_K_M with 8k of context, so the ceiling you will meet is context length and batch size, not parameter count.
Buy or rent
Rent. The purchase question does not arise outside a rack, and if it does, the answer involves cooling and power that a home does not have.
There is no purchase price to reason about, only an hourly one, and it differs by provider and by week. Put the figure you are actually looking at, with your electricity rate and your real monthly hours, into the build vs rent calculator. 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.
Rented, an A100 80 GB (PCIe or SXM) costs from $1.35 an hour (Hyperstack), on demand, before tax, read 19 October 2026. Every provider's price is in the cloud GPU price guide.
The honest take
The old datacenter workhorse, still fast where it counts, and cheap to rent.
The A100 is two generations old and still has 2,039 GB/s of HBM2e, which is more than double a 4090. For running a large model in one conversation, it remains excellent, and because newer hardware has moved the market on, it is usually one of the cheapest ways to rent real capacity. It is not the card to pick for training at scale any more.
What it is good at
- 80 GB and 2,039 GB/s: 70B models at Q4 run quickly with long context.
- Mature software support; nothing about it surprises a runtime.
- Generally the cheapest 80 GB rental class available.
What it is not
- Ampere-generation compute; slower than an H100 on prompt processing and on training.
- No FP8 support, which newer runtimes increasingly assume.
- Rental only, in practice. This is not a card you put in a desktop.
The thing people get wrong: Its memory bandwidth still beats every consumer card except the 5090 and RTX PRO 6000, and it launched in 2020. Bandwidth ages more slowly than compute, which is why old datacenter cards stay useful for chat long after they stop being bought for training.
Compared with the alternatives
- H100 SXM: 80 GB · 3,350 GB/s · best fit Llama 3.3 70B at ~57 tok/s. The fastest memory here by a wide margin, and almost always more than one person needs.
- L40S: 48 GB · 864 GB/s · best fit Muse Glimmer 30B at ~35 tok/s. The 48 GB card you meet as a rental line item, not as a purchase.
- RTX PRO 6000 Blackwell: 96 GB · 1,792 GB/s · best fit Llama 3.3 70B at ~31 tok/s. Ninety-six gigabytes and 5090-class bandwidth on one card.
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 A100 run a 70B model?
- Yes. Llama 3.3 70B at Q4_K_M with 8k of context needs about 45.8 GB, inside the 80 GB this card can address, and it generates an estimated 35 tokens per second, which is comfortable for chat.
- What is the best model to run on an A100?
- Llama 3.3 70B. At Q4_K_M and 8k context it needs about 45.8 GB of the 80 GB available and generates an estimated 35 tokens per second, which is comfortable for chat. If quality matters more than size, Muse Glimmer 30B fits at Q8 in about 33.5 GB.
- How many tokens per second does an A100 generate?
- It depends entirely on the size of the model, because generating a token means reading every active weight out of memory. At 2,039 GB/s and 70% efficiency, this card produces an estimated 308 tokens per second on an 8B model at Q4, and about 379 on the largest model it holds, Mistral Small 4 119B (MoE). Reading speed is roughly 5 to 8 words per second, so anything above 25 feels immediate. Treat 308 as a ceiling on 2,039 GB/s rather than a measurement; the higher a figure, the further real runtimes fall below it.
- Is 80 GB enough for running LLMs locally?
- It runs 34 of the 34 models tracked here at Q4_K_M with 8k of context, up to 119B parameters. The honest test is not the model list but the context: Llama 3.3 70B on this card holds about 100,352 tokens before memory runs out.
- A100 or H100 SXM for local models?
- Both address about the same memory, so they run the same models. On speed, the H100 SXM is faster: 3,350 GB/s against 2,039 GB/s, and bandwidth is what sets chat speed.
- Is an A100 still worth renting in 2026?
- For single-user inference, yes, and often it is the sensible choice. Its 2,039 GB/s of memory bandwidth still beats every consumer card except the 5090 and the RTX PRO 6000, and because newer hardware has moved the market on, it is usually the cheapest way to rent 80 GB. What it lacks is FP8 and the training throughput of an H100.
- A100 40 GB or 80 GB?
- The 80 GB version is the one that matters for language models: it holds a 70B model at Q4 with a long context, where the 40 GB card is limited to roughly the same tier as a 48 GB workstation card. It also has higher memory bandwidth.
- 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 23 of 34 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): 80 GB, 2,039 GB/s; see NVIDIA's own page for this card.
- Hourly prices: each provider's own pricing page or price API, read 19 October 2026, as listed in the cloud GPU price guide.
- Model architecture values come from each model's published config; the self-hosted LLM guide explains what each one changes.