Twoody
Local LLM calculator
The best local LLM for your NVIDIA graphics card
From the RTX 3060 to the RTX 5090: the strongest model that fits each card's memory, and how fast it writes.
| Your computer | Memory | Strongest that runs comfortably | Writes | Suggested by Twoody by default |
|---|---|---|---|---|
GeForce RTX 5090 (32 GB) |
32 GB | Qwen3.6 35B-A3B One click in Twoody | 100+ tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 5080 (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 66–95 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 5070 Ti (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 63–91 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 5070 (12 GB) |
12 GB | Gemma 4 12B One click in Twoody | 56–81 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 5060 Ti (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 32–46 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 5060 Ti (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 77–154 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 5060 (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 82–147 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 4090 (24 GB) |
24 GB | Qwen3.6 35B-A3B One click in Twoody | 100+ tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4080 SUPER (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 52–74 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4080 (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 50–72 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4070 Ti SUPER (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 47–67 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4070 Ti (12 GB) |
12 GB | Gemma 4 12B One click in Twoody | 47–67 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4070 SUPER (12 GB) |
12 GB | Gemma 4 12B One click in Twoody | 47–67 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4070 (12 GB) |
12 GB | Gemma 4 12B One click in Twoody | 41–59 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4060 Ti (16 GB) |
16 GB | Qwen3 14B One click in Twoody | 21–30 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 4060 Ti (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 53–105 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 4060 (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 53–95 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 3090 (24 GB) |
24 GB | Qwen3.6 35B-A3B One click in Twoody | 100+ tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 3080 (10 GB) |
10 GB | Gemma 4 12B One click in Twoody | 62–89 tokens/s | Qwen3.5 9B (7.9 GB) |
GeForce RTX 3070 (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 62–123 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 3060 Ti (8 GB) |
8 GB | Qwen3.5 4B One click in Twoody | 70–140 tokens/s | Qwen3.5 9B (7.9 GB) Partly on the processor: much slower |
GeForce RTX 3060 (12 GB) |
12 GB | Gemma 4 12B One click in Twoody | 33–48 tokens/s | Qwen3.5 9B (7.9 GB) |
Speeds are estimates, not measurements made with Twoody, computed from public llama.cpp benchmarks. Figures reviewed on October 16, 2026.
How to read this table
One line per graphics card: its own memory decides which models fit, and the speeds are those of a model running on the card. The rest of the PC is assumed to have 32 GB of memory.
- Strongest that runs comfortably
- The strongest model that fits comfortably there and writes at least at your reading speed, 5 tokens per second; under its name, whether it installs in one click in Twoody.
- Writes
- Its speed there, in tokens per second: an estimate from public llama.cpp benchmarks, not a measurement made with Twoody.
- Suggested by Twoody by default
- The model the app installs first on that computer, and the memory it needs there. Not always the strongest: one Twoody has measured with its memory, its tools and your documents.
Twoody is in private beta: leave your email to request access.
On a PC, the graphics card's memory (VRAM) decides: 8 GB for models of 4 to 8 billion parameters, 16 GB for 14 billion or gpt-oss 20B, 24 to 32 GB for 32 billion.
Graphics cards read their memory very fast: with a model that fits, they write faster than any Mac.
Twoody for Windows and Linux is in private beta: on Windows 10 and 11 (x64), it runs the model on an NVIDIA card through Vulkan when the driver supports it, and on the processor otherwise; on Linux, on the processor. The speeds on this page are estimates for a model running on the graphics card, not measurements made with Twoody.
Questions about running an LLM locally
Which LLM for an RTX 4090?
The strongest model that runs comfortably there is Qwen3.6 35B-A3B, which writes 100+ tokens/s. It installs in one click from Twoody's catalogue. On a PC with 32 GB of memory, Twoody suggests Qwen3.5 9B by default, which needs 7.9 GB there: one of the models it has measured with its memory, its tools and your documents.
Is an 8 GB graphics card enough?
For small models, yes. The strongest model that runs comfortably there is Qwen3.5 4B, which writes 53–154 tokens/s depending on the card. It installs in one click from Twoody's catalogue. Larger ones spill into the computer's memory and slow down a lot.
Does Twoody run on Windows with an NVIDIA card?
In the private beta, yes: on Windows 10 and 11 (x64), Twoody runs the model on an NVIDIA card through Vulkan when the driver supports it, and on the processor otherwise. Leave your email on the platforms page to join the waitlist.
The best local LLM, by computer
Try Twoody before it opens to everyone.
The private beta is free, on Mac, Windows and Linux, and on iPhone and Android with Twoody on your computer. Leave your email to ask for an invitation: we will write to you when it is ready for you.