Skip to content

Wire Briefings Search

Fetching from the wire…

01

02

03

04

05

06

07

08

Wire Briefings Search

Entity trail

Ternary Bonsai

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

1

Findings

1

Edges

3

Sources

3

Corpus findings

2026-04-17 / reddit-researcherPrismML Launches Ternary Bonsai 1.58-Bit Models — Then Community Contests Claims vs Gemma 4PrismML released Ternary Bonsai, a family of 1.58-bit ternary ({-1,0,+1}) language models in 8B, 4B, and 1.7B sizes under Apache 2.0. The 8B variant uses ~1.75GB RAM (9x less than a standard 16-bit 8B model) and claims within 5% of Qwen 3 8B performance. The announcement hit 331↑ (81 comments) on r/LocalLLaMA. However, a counter-post at 124↑ (54 comments) titled 'Bonsai models are pure hype: Bonsai-8B is MUCH dumber than Gemma-4-E2B' directly challenges the claims with benchmark comparisons.

Graph relationships

RELEASED
PrismML
PrismML released Ternary Bonsai family of 1.58-bit ternary models.
Source finding
BENCHMARKED AGAINST
Gemma 4
Community contests Ternary Bonsai performance claims versus Gemma 4.
Source finding
BENCHMARKED AGAINST
Qwen 3
Ternary Bonsai claims within 5 percent of Qwen 3 8B performance.
Source finding

Source trail

GPT-Rosalind PI's latest model Ternary Bonsai

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Ternary Bonsai intelligence trail | MindPattern