Investors
LLM compression for private and on-device deployment.
The problem, product, benchmark evidence, milestones, and current fundraising stage for Minutus AI.
The problem
On-device and private AI is limited by memory. The best model that fits the hardware wins, and standard compression degrades sharply below 4 bits.
What we do
Post-training compression of any open or fine-tuned language model to 2-3 bits per weight, without retraining, tuned to the customer's workload.
Proof so far
- Qwen3-8B compressed to 3.90 GB: 4.2x smaller than the 16.4 GB original, with the same instruction-following score (82.1).
- Matches Qwen's official 4-bit release (6.10 GB) in 36% less memory: instruction following 82.1 vs 80.6 and coding 75.0 vs 74.4 (differences not statistically significant); the 4-bit release is 1 point ahead on knowledge (71.9 vs 70.9).
- A compressed 8B in 3.90 GB scores like the uncompressed 4B in 8.04 GB: half the memory for the larger model.
- Qwen3-4B at 3 bits matches standard 4-bit quality on all three benchmarks (instruction following 78.0 vs 77.8, coding 72.6 vs 73.2, knowledge 66.3 vs 66.8; no significant difference) in less memory (2.45 GB vs 2.65 GB).
- Apertus-8B at 3 bits keeps 89% of its benchmark quality on average. Long-context ability is preserved: 95-98% of the original up to 16k tokens.
Where we are
Research pipeline validated on two model families (Qwen3, Apertus). Pilots: in discussion with first design partners. Next milestones: production runtime, measured memory on device, more model families, first paying pilots.
Funding
We are raising a pre-seed round.
Minutus AI · Zürich, Switzerland