Local ultra-dense 27B quant showdown - The results are in:
This is a AI post classified by Jev as Open & local models (a model release), kept by the AI Radar because it carries real work, not commentary.
Local ultra-dense 27B quant showdown - The results are in: • Ternary Bonsai 27B PQ2_0 runs comfortably on my 3080 10GB GPU, including up to over 50k of context. At 20k, I see 56 tok/s. • Qwen3.8 Unleashed IQ2_S slows down quickly: 36 tok/s at 8K, 25.5 at 20K, and 8.4 at 32K. Its 50K-context runs effectively stalled. • In terms of capability: Both scored strong on tool calls and context search. Bonsai has an edge with my 'end-to-end' agentic loops however, and the ability to work with more context (50k+) takes it over the top. These rebuilds from Eric are a great service to the community -
Posted by GooGZ AI (1.4k followers) 1 h ago · 1 likes · 101 views · view the original post on X. Kept by the AI Radar as Open & local models.
More AI work like this
- きしださんのjevを採用させていただいています — @Tono_Ken3
- Hermes local models nailed this habit right for me. Every catalog row shows green /… — @MichaelGannotti
- なぜこんなマイナーなのにいい感じのModelをHFからサーベイすることが出来るんだい、ブラザー。 — @studio_yebisu
- They told us the models were going away. So Pirate Face built the continuation plan. — @thepirateface
- Bonsai 2 27B PTQ1_0 is working on A20 Pro (18PM) — @imnotchalk
- A 27B AI model running locally inside a browser.🚀 — @BkashJosi
- 本地跑的开源 AI，已经进化到这种地步了吗……🤯 — @mylifcc
- Deepseek v4.1 hillclimb hit 16.58 tps after switching to Fable. So far every Q2 attempt… — @TheDavidTai
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 34.3k posts from 4.8k X accounts over the last 14 days, 1.4k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-20 12:51 UTC. Full method.