llama.cpp weekly picks: Liquid LFM3, 1T-param Mistral Large 4, and Flint hit 74.2% SWE-bench
mervenoyann · x · 2026-10-31
llama.cpp's weekly model picks, all runnable locally in llama.app:
- Liquid AI LFM3 family: LFM3-1.7B (16-language text, CPU/GPU), LFM3-Omni-2B (see/hear/speak omni model), LFM3-Audio-2B, and d1-2B decision model outputting yes/no, choices and scores across text/image/video.
- Mistral Large 4: 1T params (52B active), 1M context, image input; the 2-bit quant fits in 512GB unified memory, strong at cybersecurity and agentic coding.
- Recursive Flint: 2.5B active of 39.8B coding-agent model scoring 74.2% on SWE-bench Verified, 262K context, built for codebase investigation, bug fixing and test writing.
More from Models
- Arena weekly: Gemini 4 Carbon tops Text Arena, GPT-6.1 Terra enters Agent Arena at #8 — arena · 2026-10-31
- 6 October models, same 5 trap-laden agent tasks: DeepSeek does it all for $0.02, hyped Flint passes 1/5 — latent_lurker8587 · 2026-10-31
- Security researcher: air-gapped AI agents are only possible with open-weight models — wunderwuzzi23 · 2026-10-31
- Gemini 4 Carbon tops Artificial Analysis index at 59, 23% cheaper per task than Opus 5.5 — rohanpaul_ai · 2026-10-31
- Podcast: Is Mistral Large 4 the West's best open model? Plus AI cyber threats — The Cognitive Revolution · 2026-10-31
- OpenAI shows GPT-6 Astra remotely controlling a robot arm in builder community — OpenAIDevs · 2026-10-31