Startup profile · funding coverage
PrismML funding, valuation and investors
Caltech-founded company building compressed 1-bit and ternary Bonsai models for consumer and edge hardware.
Verified company research · funding, investors and latest developments
Keep track of PrismML
Save this profile to your VCT watchlist for a quick return.
Funding, valuation & investors
Answer-first snapshotLatest funding
Seed
$22.25M · September 17, 2026
Latest known valuation
Not publicly disclosed
Total disclosed equity funding
$22M
Excludes debt, grants, acquisitions, secondaries, and IPO proceeds.
Current status
private
seed · Pasadena, California, United States
Investors in latest funding
Other: Khosla Ventures , Cerberus Ventures , California Institute of Technology (Caltech)
Overview
PrismML is a Pasadena-area AI company founded by Caltech researchers and led by Babak Hassibi. It develops Bonsai 1-bit and ternary language models that compress model weights for on-device inference. TechCrunch reported a $22.25M seed by September 17, 2026, backed by Khosla Ventures, Cerberus and Caltech.
Financial databases place the round earlier in 2026, with disagreement over the exact closing date, so VCT uses the public-report date without claiming it was the legal closing date.
Why PrismML is interesting
A genuine edge-AI efficiency financing with $22.25M in seed funding. PrismML reduces model weight storage and hardware requirements, not the autoregressive token-generation bottleneck addressed by NCM.
Product & use cases
Bonsai small-footprint language and image models using extremely low-precision neural weights, targeting smartphones, PCs and edge devices.
- On-device private language generation
- AI inference on memory-limited phones, PCs and embedded devices
- Data-center model compression and efficiency
Key facts
- $22.25M seed publicly reported on September 17, 2026
- Khosla Ventures, Cerberus and Caltech named as investors; lead not publicly resolved
- Exact seed legal close date disputed across financial databases; reporting date used
- Bonsai model family targets lower-weight memory needs on consumer hardware
Funding history (newest first)
Seed
2026-09-17 (reporting date; closing date unverified) $22.25M- Khosla Ventures (participant)
- Cerberus Ventures (participant)
- California Institute of Technology (Caltech) (participant)
Source: Read source · techcrunch.com
Competitive landscape
Edge: Low-bit model design from Caltech researchers and models released publicly; benchmark quality and device speed depend on quantization and implementation.
PrismML's memory savings do not themselves demonstrate lower end-to-end inference cost or frontier generalization. It competes with native low-bit training and post-training compression; not an NCM/diffusion-model startup.
Market / IPO context
Editorial / static context — not a live quote.
FAQs about PrismML
Practical answers founders, operators, and investors typically search for.
Last updated:
Editorial note: AI tools assisted with research, structure, or drafting. Venture Capital Tracker retains human editorial responsibility for factual accuracy, relevance, and source quality before publication.