Trapetum

Press the model, keep the oil.

Run real LLMs at 4-bit on your own GPU. Pure Rust, no Python at runtime, nothing leaves your machine. Trapetum presses the weights into a small 4-bit code and decodes it straight inside the matmul, so the full weight matrix is never materialized. (The name is the Roman olive press: crush the fruit, keep the essence.)

Run it on your machine View on GitHub

See the benchmarks · Read the paper

16/16

tokens match HuggingFace

135

tok/s on one RTX 4090

3.5 GB

weights, vs 13 GB fp16

2.1×

less energy per token

Real Llama-2-7B, 4-bit, in pure Rust. No Python at runtime.

What do these numbers mean?

16/16 tokens match HuggingFace: given the same prompt, the 4-bit model emits the exact same next 16 tokens (greedy) as the original fp16 model. Proof the quantization, the kernel and the Rust pipeline are correct, not approximate (worst-case logit gap 7.9e-3).
135 tok/s on one RTX 4090: generation speed, one token at a time (batch 1), on a single consumer GPU. How fast it “types” the answer.
3.5 GB on disk (4.7 GB VRAM), vs ~13.5 GB fp16: the weights shrink from ~13 GB (16-bit) to 3.5 GB (4-bit), so a 7B model fits on a small or consumer GPU instead of needing a datacenter card.
2.1× less energy per token: each generated token costs 2.1× less energy than fp16 (2.58 vs 5.45 joules per token, measured from GPU power draw). Cheaper and greener to run.

NVIDIA CUDA

Runs on any modern NVIDIA GPU, not just the 4090

The RTX 4090 above is only our reference card. Trapetum is CUDA and runs on any Ampere, Ada or Hopper GPU (A100, A40, RTX 30/40, H100, and more): recompile with one -arch flag, or ship a single multi-arch binary. The 4-bit memory win is universal (a 7B in ~3.5 GB fits an 8 GB card); the decode speedup follows the bandwidth law, largest on consumer cards, parity on an H100. See the per-GPU table.

It reproduces HuggingFace, token for token

Loaded from a 3.5 GB .cbk file, the runtime decodes the same quantized weights as HuggingFace and reproduces its greedy generation exactly, with no Python in the loop.

$ ./trapetum llama-2-7b.cbk          # pure Rust, no Python

loaded  llama-2-7b.cbk  (3.5 GB)  in 3.1 s
prompt:  "The capital of France is"

logits vs HuggingFace (worst)    7.9e-3
top-1 agreement with HF          6 / 6
greedy continuation = HF        16 / 16 tokens   OK
decode throughput               135 tok/s
energy                          2.58 J/token   (2.1x less than fp16)

An honest Pareto, one machine

Speed, memory, accuracy and energy for the same Llama-2-7B on an RTX 4090, batch-1 decode, iso-context. The pure-Rust runtime is the fastest and the lowest-energy path.

Trapetum vs baselines: decode speed and energy on Llama-2-7B — fp16 vs Trapetum 4-bit (Python and Rust) vs AQLM 2-bit. Rust runtime in orange.

Honest limits

It does not beat dense fp16 on accuracy. Nothing does; quantization trades a little perplexity for a lot of memory.
The kernel speedup is a memory-bandwidth law: largest on consumer GPUs, near parity on an H100 where fp16 is already near roofline.
Batch-1 decode for now. Batched throughput is an open problem, marked as such.

Everything is public, measured on real hardware, and reproducible with one command, including the negative results.

github.com/neuralboot/trapetum