LiquidAI Launches LFM2.5-2.6B: An On-Device Agentic Model
JosephJacks_ · x · 2026-08-12
LiquidAI released LFM2.5-2.6B, an agentic model designed to run entirely on-device across phones, laptops, and robots. It supports planning, tool calling, and multi-step tasks while keeping data local.
Pre-trained on 34T tokens with a hybrid architecture, the model supports a 128K context length. Official benchmarks show it outperforming larger models like Qwen3.5-9B in ToolSandbox. It is currently available for testing in their local AI app, Apollo.
More from Models
- Users Accuse Anthropic of Cooked Evals, Claiming Real API Performance Lags — GabGarrett · 2026-08-12
- Encrypted Chain-of-Thought in Proprietary LLMs Can Be Extracted via Weaker Sibling Models — Simon Willison · 2026-08-12
- GPT vs Claude: Which Model Better Understands Your Intent? — nickbaumann_ · 2026-08-12
- Expert Warns AI Watermarks Could Increase Code Entropy, Harming Coding Agents — wightmanr · 2026-08-12
- e/acc's Beff Jezos flags a 'vibe shift' as more users daily drive Grok — beffjezos · 2026-08-12
- Kimi K3 Distillation Controversy: Authors Admit No Proof, Likely Data Contamination — bookwormengr · 2026-08-12