Skip to main content
Back to News
PrismML's Ternary Bonsai 2 27B compresses a 27-billion-parameter model to under 6GB while retaining 98.2% of baseline performance.
Technology
2 min read
US

PrismML's Ternary Bonsai 2 27B compresses a 27-billion-parameter model to under 6GB while retaining 98.2% of baseline performance.

The AMW Read

Second-generation ternary quantization from a known Apple-acquisition-rumored player meaningfully improves benchmark retention (98.2% vs ~95%) via an explicit open-weight release strategy.
NoveltySignificance
Foundation Models Β· Player MapScaling Laws
PrismML
PrismML

Foundation Models / LLMs

View Company Profile

PrismML's Ternary Bonsai 2 27B compresses a 27-billion-parameter model to under 6GB while retaining 98.2% of baseline performance.

PrismML, a Caltech spinoff previously linked to Apple acquisition talks after shrinking a 54GB Qwen model to under 4GB for on-device use on the iPhone 17 Pro in July, released Ternary Bonsai 2 27B on September 17. Built on Alibaba's open-source Qwen2.5 27B, the model replaces standard 16-bit weights with a ternary scheme using only three values (-1, 0, +1) combined with FP16 group-level scaling, shrinking the 53.8GB FP16 original to 5.93GB β€” about 1.76 effective bits per parameter. Across 20 benchmarks it retained 98.2% of the original model's performance, up from roughly 95% for the first Bonsai 27B two months earlier, though it still lags on multi-step agentic work: 52.8 versus 69.7 on Terminal-Bench 2.1, and 60.8 versus 80.6 on SWE-bench Verified.

The release sits in the on-device AI compression race, where the binding constraint is whether a capable model fits in a phone's or laptop's memory rather than raw capability. PrismML's ternary weighting pushes compression further than typical INT4/INT8 quantization while narrowing the accuracy gap that has usually made aggressive compression unusable outside toy benchmarks. Multimodal input, a 262K-token context window, and native CUDA and Apple MLX execution paths target both edge GPU and consumer-device deployment β€” extending, not just repeating, the iPhone demo that first drew Apple's interest.

A 27B-class model under 6GB running at 142.5 tokens/second on an RTX 5090 and 46.8 on an Apple M5 Max makes locally run coding assistants and document or computer-control agents workable without server round-trips, provided the task avoids long agentic chains, where the compression penalty persists. The Apache 2.0 release on Hugging Face, paired with standing Apple acquisition speculation, points to a company monetizing licensing interest rather than API revenue; per the AI Market Watch index, the Caltech spinoff has raised $16.25M to date (coverage, not a census) β€” a modest base for a startup now positioned as compression IP a hyperscaler might want to own outright.

#PrismML #OnDeviceAI #ModelCompression #Quantization #Qwen #EdgeAI

#PrismML#model compression#ternary quantization#on-device AI#related:Apple

How This Connects

Based on Foundation Models Β· Player Map

  1. 1d agoThinking Machines releases Inkling 2 Mini for local use on laptopsThinking Machines
  2. 1d agoOpenAI reports chip design gains on Jalapeno and AMD hardwareOpenAI
  3. 6d agoTranslated releases Lara Mini Base, a multilingual 7.2B modelTranslated
  4. 1w agoDeepSeek Reportedly Raises $15 Billion With Tencent and CATL ParticipationDeepSeek
  5. 1mo agoReka releases Flint for local coding, citing AI-led model developmentReka
  6. 2mo agoPrismML's Ternary Bonsai 2 27B compresses a 27-billion-parameter model to under 6GB while retaining 98.2% of baseline performance. Β· THIS ARTICLE

Related News

More news from PrismML

Stay updated with the latest news and announcements from PrismML.

View all PrismML news

Discover AI Startups

Explore 5,000+ AI companies with VC-grade analysis, funding data, and investment insights.

Explore Dashboard