https://aiagentsnews.top/tags/models/ Recent content in Models on AI Agents News Hugo en Sat, 25 Jul 2026 12:45:21 +0000 https://aiagentsnews.top/posts/mixture-of-agents-why-hermes-06-beats-single-models/ Sun, 19 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/mixture-of-agents-why-hermes-06-beats-single-models/ Hermes 0.6 uses parallel reference models to isolate reasoning from tool use, preventing hallucinated actions while improving decision quality. https://aiagentsnews.top/posts/inference-latency-why-41b-active-beats-975b/ Thu, 16 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/inference-latency-why-41b-active-beats-975b/ Learn why 41B active parameters matter more than total count for agent latency. Discover selection criteria for reliable, cost-effective systems. https://aiagentsnews.top/posts/agent-runtime-rules-boost-scores-by-14-points/ Tue, 14 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/agent-runtime-rules-boost-scores-by-14-points/ See how SelfUse jumped TerminalBench scores from 23.8% to 38.1% by mining execution traces instead of tweaking model weights manually. https://aiagentsnews.top/posts/langchain-custom-tools-stop-llm-guessing-now/ Tue, 14 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/langchain-custom-tools-stop-llm-guessing-now/ Learn to build a CircumferenceTool in LangChain. This guide shows how custom tools replace token-based guessing with grounded, repeatable functions. https://aiagentsnews.top/posts/native-interaction-models-why-04s-latency-matters/ Tue, 14 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/native-interaction-models-why-04s-latency-matters/ Thinking Machines' native interaction models hit 0.40s latency, outpacing GPT-realtime2's 1.18s for concurrent audio and text streams. https://aiagentsnews.top/posts/orchestrator-worker-pattern-cuts-ai-costs-10x/ Tue, 14 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/orchestrator-worker-pattern-cuts-ai-costs-10x/ Splitting orchestration from execution can cut inference costs by around 10x. Learn how the orchestrator-worker pattern separates planning from routine tasks. https://aiagentsnews.top/posts/agentic-ai-2026-free-tools-that-execute-tasks/ Wed, 08 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/agentic-ai-2026-free-tools-that-execute-tasks/ Perplexity removed Comet's paywall in March 2026. Explore free agentic tools that decompose goals and execute sequences without human prompts. https://aiagentsnews.top/posts/terminalbench-v21-89-tasks-for-real-agent-mastery/ Tue, 07 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/terminalbench-v21-89-tasks-for-real-agent-mastery/ TerminalBench v2.1 uses 89 curated tasks to test if AI agents can execute complex system commands rather than just generating static code snippets. https://aiagentsnews.top/posts/agent-frameworks-compared-langgraph-02-vs-crewai-086/ Mon, 06 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/agent-frameworks-compared-langgraph-02-vs-crewai-086/ Gartner predicts 40% of enterprise apps will use task-specific agents by 2027. Compare how LangGraph and CrewAI handle this complex shift. https://aiagentsnews.top/posts/hermes-agent-models-sonnet-46-vs-deepseek-v4/ Sun, 05 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/hermes-agent-models-sonnet-46-vs-deepseek-v4/ Compare Claude Sonnet 4.6 ($3) against DeepSeek V4 ($0.30) for Hermes Agent. Learn which model prevents malformed tool calls in production. https://aiagentsnews.top/posts/react-agents-vs-function-calling-2026-guide/ Sun, 05 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/react-agents-vs-function-calling-2026-guide/ Compare ReAct's iterative reasoning loops against structured function calling for external API access in 2026 agent architectures. https://aiagentsnews.top/posts/agent-tools-that-enable-action-not-just-chat/ Sat, 04 Jul 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/agent-tools-that-enable-action-not-just-chat/ Late 2025 data shows action-enabling tools are now the majority use case. Learn how nine specific tool categories prevent context window overload. https://aiagentsnews.top/posts/local-coding-agents-build-free-private-stacks/ Mon, 29 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/local-coding-agents-build-free-private-stacks/ Skip the $7.60 per task fee. I show how to run Qwen3.6 locally on 32GB RAM for private, zero-cost code generation. https://aiagentsnews.top/posts/terminalbench-21-exposes-agent-wrapper-gaps/ Sun, 28 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/terminalbench-21-exposes-agent-wrapper-gaps/ Codex CLI paired with GPT-5.5 sits at the top of the Terminal-Bench 2.1 leaderboard with an 83.4% pass rate. https://aiagentsnews.top/posts/llamaindex-agents-connect-live-data-sources/ Fri, 26 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/llamaindex-agents-connect-live-data-sources/ Learn how LlamaIndex uses over 200 data loaders to connect static LLMs to live enterprise sources and bypass training cutoffs. https://aiagentsnews.top/posts/glm52-performance-rivals-closed-coding-models/ Tue, 23 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/glm52-performance-rivals-closed-coding-models/ GLM-5.2 improved internal task success rates from 21/70 to 48/70 over its predecessor, signaling a shift in open-weight viability. https://aiagentsnews.top/posts/recursive-improvement-risks-anthropics-2026-code-surge/ Mon, 15 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/recursive-improvement-risks-anthropics-2026-code-surge/ Anthropic reports an 80-fold revenue surge in 2026 as recursive self-improvement accelerates, raising urgent questions about frontier safety protocols. https://aiagentsnews.top/posts/mai-models-microsofts-35b-reasoning-engine-explained/ Mon, 15 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/mai-models-microsofts-35b-reasoning-engine-explained/ Microsoft's MAI-Thinking1 uses 35B active parameters within a 1T-parameter MoE to cut costs while handling 256K context windows for complex reasoning. https://aiagentsnews.top/posts/ollama-models-stop-ai-ghostwriting-on-windows/ Mon, 15 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/ollama-models-stop-ai-ghostwriting-on-windows/ Shift from ghostwriting to coaching with Ollama on Windows. This guide details a 100% offline workflow that keeps sensitive data secure. https://aiagentsnews.top/posts/prompt-patterns-beat-trial-and-error-for-llms/ Mon, 15 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/prompt-patterns-beat-trial-and-error-for-llms/ Stop iterative guessing by applying a strict 5-block prompt architecture that shifts probabilistic model outputs toward accurate, stable results. https://aiagentsnews.top/posts/small-llm-on-intel-i5-36-tokenssecond-reality/ Mon, 15 Jun 2026 00:00:00 +0000 AI Agents News Editorial https://aiagentsnews.top/posts/small-llm-on-intel-i5-36-tokenssecond-reality/ Testing five models on an Intel i5 reveals 36 tokens per second is the ceiling for small LLMs due to 20 GB/s memory bandwidth limits.