Unsloth release runs 26.4B Inkling 2 Mini on 16GB RAM, cuts MCP token use 87%
danielhanchen · x · 2026-11-21
Unsloth shipped v0.1.922-beta, adding support for Thinking Machines Lab's Inkling 2 Mini — a 26.4B (5.4B active) model with 1M context that the team calls the strongest model runnable on 16GB RAM.
Highlights:
- Faster image generation: MiniMax-H3-Image renders a new prompt in 36.7s on 8GB GPUs (down from 82.7s); GGUF image models load with 52-66% less RAM; FLUX.1 steps are 16-28% faster when weights stream from system RAM
- Mac video restyle: Apple Silicon can re-render whole clips in a new style
- MCP chats: tool conversations no longer reprocess the full history, cutting prompt tokens by 83-87% in tests
- Scanned PDFs are now OCR-read in chat; fixes include GTX 16xx NaN render errors
Docs and matching llama.cpp releases were updated alongside.
More from coding & agent
- OSU NLP: computer-use and robot-use agents started together and will converge in 2026 — taoyds · 2026-11-21
- Inkling 2 Mini builds a flight simulator fully offline on a Mac mid-flight — miramurati · 2026-11-21
- Seroter Daily Reading List #914: harnesses are cheap to swap, model behavior isn't — rseroter · 2026-11-21
- Modified ARTEX agent pentest framework stripped of programmatic scope enforcement — AccBalanced · 2026-11-21
- Dev ships mercadopago-mcp: a read-only-by-default MCP server for Mercado Pago — amarggos · 2026-11-21
- Vibe-coding culture: everyone builds over the holidays, few realize when to log off — haydendevs · 2026-11-21