244 stories
Agentic AI systems operating over geospatial data require more than autonomous reasoning and tool use; they must also support explicit orchestration, spatial validation, governance, provenance, failure handling, and human decision authority.
A practical guide to running Liquid AI's LFM2.5-24B-A2B mixture-of-experts model locally via Transformers, llama.cpp, MLX, and vLLM.
LFM2.5-24B-A2B edges out larger MoE rivals on agentic and tool-use benchmarks while using far fewer active parameters per token.
LFM2.5-24B-A2B is a 24B-parameter reasoning model with just 2B active parameters, built for fast, private agents on-device. It thinks before it answers, calls tools, and works through multi-step tasks inside popular agent harnesses.
Open models like GLM-5 and MiniMax M2.7 now match closed frontier models on core agent tasks — file operations, tool use, and instruction following — at a fraction of the cost and latency. Here's what our evals show and how to start using them in Deep Agents.
A beginner's guide to voice AI that follows one phone call from first ring to reply and ends with a 20-term glossary grouped by where each term shows up.
How cascade and speech-to-speech voice agents compare on tool calling, latency, voice and cost in 2026 benchmarks, and how to choose per agent.
Agentic workflows sit between fixed automations and fully autonomous agents. Learn what they are, the patterns behind them, real examples, and a 5-question checklist for when to use one.
Hands-on local test of MiniCPM-V 4.7, an Apache 2.0 vision MoE model, covering OCR, SQL debugging, physics reasoning and multilingual translation.
GLM 5.2 vs Sonnet 4.6 on seven real agent tasks, updated for GLM 5.3 and Sonnet 5 with verified prices and a GLM 5.1 lineage note.
Qwen3.8 27B tool calls hang in Ollama, Hermes or Claude Code? Fixed in Ollama v0.33.0. The real cause, how to confirm it, and what to do if it persists.
MiniCPM-V 4.7's benchmark results against Qwen 3.6 and Gemma, showing where the Apache 2.0 vision MoE model wins, ties, and falls short.
Underdog Saluki 27B compresses Qwen3.8-27B to under 8GB and still beats the full model on BFCL tool-calling benchmarks. Here's the data.
Searching for Gemma 4 27B or Qwen 4 27B? Neither exists yet. The real matchup is Gemma 4 31B vs Qwen 3.8 27B: tool calling, code, speed and VRAM on a 24GB card.
Seven free or near-free LLMs ranked for AI agents: Gemini 3.8 Flash, Qwen 3.8 27B, DeepSeek V4.1 Flash, GPT-6 Luna, MiMo V2.6 Flash, Gemma 4 and GLM.
MiniMax M3 at $0.30/M vs Claude Sonnet 5 ($2), Sonnet 4.6 ($3) and Opus 5 ($5). Five agent tasks tested, the routing maths, and when the premium pays.
Agentic AI plans and acts toward a goal on its own. Plain definition, 2026 examples like OpenAI Dots and Meta Muse, and agentic AI vs AI agents vs assistants.
Ornith-1.5-35B-A3B: a 35B MoE with 3B active. Ollama run command, hardware by quant, OpenClaw setup, and benchmarks vs Qwen3.6 and Qwen3.5.
Showing 18 of 244 stories
Load more storiesGet the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
