https://freeainews.com/categories/open-source/Recent content in Open Source on Free AI NewsHugoen-usFree AI NewsSat, 22 Aug 2026 11:17:26 -0400https://freeainews.com/open-source/nvidia-nemoclaw-open-source-agent-framework-2026/Sat, 22 Aug 2026 11:17:26 -0400https://freeainews.com/open-source/nvidia-nemoclaw-open-source-agent-framework-2026/NVIDIA NemoClaw is an open-source secure AI agent framework for self-hosted tool calling. Learn its Apache 2.0 license, security model, and how it compares.https://freeainews.com/open-source/new-open-source-ai-projects-on-github-and-hugging-face-april-2026/Sat, 22 Aug 2026 11:14:50 -0400https://freeainews.com/open-source/new-open-source-ai-projects-on-github-and-hugging-face-april-2026/DeepSeek V4, Qwen 3.6, Mistral Small 4, Kimi K2-7, and Nemotron 3 Ultra lead April 2026 open-source AI on GitHub and Hugging Face.https://freeainews.com/open-source/zaya1-8b-zyphra-open-source-reasoning-2026/Sat, 22 Aug 2026 02:45:35 -0400https://freeainews.com/open-source/zaya1-8b-zyphra-open-source-reasoning-2026/Zyphra released ZAYA1-8B, an MIT-licensed Mixture-of-Experts reasoning model with 128k context, 8B total parameters, and strong local benchmarks.https://freeainews.com/open-source/zaya1-8b-moe-reasoning-model/Sat, 22 Aug 2026 02:42:58 -0400https://freeainews.com/open-source/zaya1-8b-moe-reasoning-model/Zyphra's ZAYA1-8B open Mixture-of-Experts model lands with strong math and code results. Here are benchmarks, license, and how to run it.https://freeainews.com/open-source/unsloth-studio-web-ui-local-models/Sat, 22 Aug 2026 02:40:36 -0400https://freeainews.com/open-source/unsloth-studio-web-ui-local-models/Unsloth Studio's new web UI lets developers run local AI models without cloud fees. Learn setup, supported models, and why it matters.https://freeainews.com/open-source/top-5-open-source-llms-self-host-free-2026/Sat, 22 Aug 2026 02:37:58 -0400https://freeainews.com/open-source/top-5-open-source-llms-self-host-free-2026/Compare the top 5 open-source LLMs to self-host for free in 2026, including DeepSeek V4, Qwen 3.6, Llama 4 Scout, Mistral Small 4, and Gemma 4.https://freeainews.com/open-source/state-of-open-source-on-hugging-face-spring-2026/Sat, 22 Aug 2026 02:34:38 -0400https://freeainews.com/open-source/state-of-open-source-on-hugging-face-spring-2026/Spring 2026 brings open weight models from DeepSeek, Qwen, Kimi, and Zyphra. Compare licenses, benchmarks, and free local AI tools.https://freeainews.com/open-source/running-llama-3-locally/Sat, 22 Aug 2026 02:31:40 -0400https://freeainews.com/open-source/running-llama-3-locally/Learn how to run Meta's Llama 3 8B and 70B open-weight models locally with Ollama, LM Studio, llama.cpp, and Hugging Face. No API costs.https://freeainews.com/open-source/qwenpaw-open-source-ai-assistant-2026/Sat, 22 Aug 2026 02:28:58 -0400https://freeainews.com/open-source/qwenpaw-open-source-ai-assistant-2026/QwenPaw is a free open source AI assistant with a built-in web IDE, Apache 2.0 license, 32B parameters, 128K context, and local coding tools.https://freeainews.com/open-source/qwen-3-6-apache-open-source-coding-2026/Sat, 22 Aug 2026 02:26:30 -0400https://freeainews.com/open-source/qwen-3-6-apache-open-source-coding-2026/Qwen 3.6 is a free Apache 2.0 coding model with a 128K context window and benchmark scores that beat paid AI tools. See specs, license, and setup.https://freeainews.com/open-source/openjarvis-local-personal-ai-2026/Sat, 22 Aug 2026 02:23:53 -0400https://freeainews.com/open-source/openjarvis-local-personal-ai-2026/OpenJarvis brings a free local personal AI to your own hardware in 2026. Learn what shipped, benchmarks, license, and how it compares to cloud AI.https://freeainews.com/open-source/opencode-open-source-ai-coding-agent-2026/Sat, 22 Aug 2026 02:21:42 -0400https://freeainews.com/open-source/opencode-open-source-ai-coding-agent-2026/OpenCode ships as a free open source AI coding agent in 2026. See license, benchmarks, context window, and how it compares to Qwen, Hermes, and Mellum.https://freeainews.com/open-source/openai-gpt-oss-open-weight-2026/Sat, 22 Aug 2026 02:18:45 -0400https://freeainews.com/open-source/openai-gpt-oss-open-weight-2026/OpenAI gpt-oss open-weight models in 2026 bring 8B, 20B, and 120B parameter options with 256k context. Here are benchmarks, license terms, and setup.https://freeainews.com/open-source/open-source-ai-news-june-2026-startup-edition/Sat, 22 Aug 2026 02:15:04 -0400https://freeainews.com/open-source/open-source-ai-news-june-2026-startup-edition/June 2026 open-source AI releases from DeepSeek, Zyphra, Nous and Cohere: model specs, licenses, benchmarks, and free self-hosting paths.https://freeainews.com/open-source/open-generative-ai-self-hosted-studio-2026/Sat, 22 Aug 2026 02:12:39 -0400https://freeainews.com/open-source/open-generative-ai-self-hosted-studio-2026/Open Generative AI Studio is a free self-hosted platform with 200+ open-weight models, Apache 2.0 licensing, and no per-token fees.https://freeainews.com/open-source/odysseus-self-hosted-ai-workspace-2026/Sat, 22 Aug 2026 02:08:58 -0400https://freeainews.com/open-source/odysseus-self-hosted-ai-workspace-2026/Odysseus shipped June 10, 2026 as a free, Apache 2.0 self-hosted AI workspace with agents, a 7.6B model, 128K context, and local UI.https://freeainews.com/open-source/nvidia-rtx-spark-superchip-open-source-ai/Sat, 22 Aug 2026 02:06:00 -0400https://freeainews.com/open-source/nvidia-rtx-spark-superchip-open-source-ai/NVIDIA RTX Spark Superchip brings 256GB unified memory and 405B open-model support to desktops. Learn specs, benchmarks, license implications, and how to run it.https://freeainews.com/open-source/nvidia-nemotron-3-ultra-550b-open-weight-2026/Sat, 22 Aug 2026 02:02:54 -0400https://freeainews.com/open-source/nvidia-nemotron-3-ultra-550b-open-weight-2026/NVIDIA Nemotron 3 Ultra 550B open-weight model benchmarks, license, hardware requirements, and how it compares to Llama 4, DeepSeek V4, and closed AI models.https://freeainews.com/open-source/nvidia-cosmos-3-open-source-physical-ai-2026/Sat, 22 Aug 2026 01:54:42 -0400https://freeainews.com/open-source/nvidia-cosmos-3-open-source-physical-ai-2026/NVIDIA Cosmos 3 brings open-weight 4B to 34B physical AI models for robotics and autonomous systems with Apache 2.0 and Hugging Face access.https://freeainews.com/open-source/nanobot-open-source-ai-agent-2026/Sat, 22 Aug 2026 01:48:13 -0400https://freeainews.com/open-source/nanobot-open-source-ai-agent-2026/NanoBot is an open source AI agent framework with 41K GitHub stars. It runs local LLMs, supports tool use, and is free to self-host.https://freeainews.com/open-source/mistral-small-4-apache-open-source-2026/Sat, 22 Aug 2026 01:45:57 -0400https://freeainews.com/open-source/mistral-small-4-apache-open-source-2026/Mistral Small 4 delivers a free Apache 2.0 open-weight model with vision, long context, and strong benchmarks for local AI in 2026.https://freeainews.com/open-source/mistral-medium-3-5-open-weight-2026/Sat, 22 Aug 2026 01:43:14 -0400https://freeainews.com/open-source/mistral-medium-3-5-open-weight-2026/Mistral Medium 3.5 is an Apache 2.0 open-weight model with 128k context and strong coding benchmarks. See specs, how to run it, and alternatives.https://freeainews.com/open-source/ltx-23-lightricks-open-source-4k-video-2026/Sat, 22 Aug 2026 01:40:21 -0400https://freeainews.com/open-source/ltx-23-lightricks-open-source-4k-video-2026/Lightricks releases LTX-2.3, an open source 4K video generation model under Apache 2.0. Here are benchmarks, hardware requirements, and license details.https://freeainews.com/open-source/localai-43-open-source-2026/Sat, 22 Aug 2026 01:37:57 -0400https://freeainews.com/open-source/localai-43-open-source-2026/LocalAI 4.3.0 ships signed backends, making free self-hosted AI safer and easier. Here's what changed, why it matters, and how to run it.https://freeainews.com/open-source/gemma-4-open-source-2026/Sat, 22 Aug 2026 01:35:06 -0400https://freeainews.com/open-source/gemma-4-open-source-2026/Gemma 4 vs GPT-4o in 2026: compare Apache 2.0 open weights, 256K context, benchmarks, local hardware, and API cost for developers.https://freeainews.com/open-source/mistral-ai-latest-open-source-release/Sat, 22 Aug 2026 01:29:47 -0400https://freeainews.com/open-source/mistral-ai-latest-open-source-release/Mistral AI released Mistral Small 3.1 under Apache 2.0. The 24B model beats GPT-4o mini and Gemma 3 on key benchmarks and runs on a single GPU.https://freeainews.com/open-source/minimax-m3-open-source-2026/Sat, 22 Aug 2026 01:27:31 -0400https://freeainews.com/open-source/minimax-m3-open-source-2026/MiniMax M3 ships as the first open-weight model with a 1M-token context. Get specs, benchmarks, license, and self-hosting details.https://freeainews.com/open-source/microsoft-aion-10-instruct-edge-ai-2026/Sat, 22 Aug 2026 01:24:39 -0400https://freeainews.com/open-source/microsoft-aion-10-instruct-edge-ai-2026/Microsoft ships Aion 1.0 Instruct, a free open on-device model for Edge. See specs, license, benchmarks, and how it compares.https://freeainews.com/open-source/microsoft-agent-framework-open-source-2026/Sat, 22 Aug 2026 01:21:51 -0400https://freeainews.com/open-source/microsoft-agent-framework-open-source-2026/Microsoft's open-source Agent Framework SDK ships with MIT license, model-agnostic tools, guardrails, and evaluation for building AI agents.https://freeainews.com/open-source/mellum2-jetbrains-open-source-2026/Sat, 22 Aug 2026 01:19:03 -0400https://freeainews.com/open-source/mellum2-jetbrains-open-source-2026/JetBrains Mellum2 is a 12B-parameter open-source code model with 128K context and Apache 2.0 license. See benchmarks, hardware needs, and how to run it.https://freeainews.com/open-source/llamacpp-latest-releases-performance-features-support/Sat, 22 Aug 2026 01:09:47 -0400https://freeainews.com/open-source/llamacpp-latest-releases-performance-features-support/LLaMA.cpp's latest release boosts local LLM speed, adds broader GPU support, and keeps MIT license. Here's what changed and how to run it.https://freeainews.com/open-source/llama-4-scout-maverick-open-source-2026/Sat, 22 Aug 2026 01:06:42 -0400https://freeainews.com/open-source/llama-4-scout-maverick-open-source-2026/Meta Llama 4 Scout and Maverick open-source models deliver 10M context, MoE efficiency, and commercial licenses. See specs, benchmarks, and how to run them.https://freeainews.com/open-source/liquid-ai-lfm2-5-on-device-open-weight-2026/Sat, 22 Aug 2026 00:47:48 -0400https://freeainews.com/open-source/liquid-ai-lfm2-5-on-device-open-weight-2026/Liquid AI LFM2.5 is a free open-weight model family with 1.3B to 12.5B parameters, 32k context, and on-device inference for any hardware.https://freeainews.com/open-source/lerobot-huggingface-open-source/Sat, 22 Aug 2026 00:44:53 -0400https://freeainews.com/open-source/lerobot-huggingface-open-source/LeRobot ships Apache 2.0 robotics datasets, models, and training tools. Compare LeRobot vs ROS 2, Isaac Lab, and Open X-Embodiment.https://freeainews.com/open-source/kimi-ai-releases-open-source-k2-7-code-model/Sat, 22 Aug 2026 00:41:46 -0400https://freeainews.com/open-source/kimi-ai-releases-open-source-k2-7-code-model/Moonshot AI released Kimi K2.7 Code, a 1T-parameter open-source coding model with 128K context. See benchmarks, license terms, and how to run it locally.https://freeainews.com/open-source/hugging-face-ml-intern-open-source-ml-engineer-2026/Sat, 22 Aug 2026 00:38:07 -0400https://freeainews.com/open-source/hugging-face-ml-intern-open-source-ml-engineer-2026/Hugging Face ml-intern is a free Apache 2.0 ML engineer agent with 7B parameters, 32k context, and local deployment. See benchmarks and setup.https://freeainews.com/open-source/hugging-face-ml-intern-open-source-llm-post-training-agent-2026/Sat, 22 Aug 2026 00:35:00 -0400https://freeainews.com/open-source/hugging-face-ml-intern-open-source-llm-post-training-agent-2026/Hugging Face ml-intern is a new open-source LLM training agent. It ships a 7B controller, 32k context, Apache 2.0 license, and agentic fine-tuning automation.https://freeainews.com/open-source/hugging-face-free-inference-gemma-4/Sat, 22 Aug 2026 00:31:35 -0400https://freeainews.com/open-source/hugging-face-free-inference-gemma-4/Google and Hugging Face shipped Gemma 4 on June 10, 2026. Free multimodal image and text inference is here. Compare free access tiers.https://freeainews.com/open-source/hugging-face-brings-open-source-llms-to-github-copilot-chat-in-vs-code/Sat, 22 Aug 2026 00:29:05 -0400https://freeainews.com/open-source/hugging-face-brings-open-source-llms-to-github-copilot-chat-in-vs-code/GitHub Copilot Chat now runs open LLMs from Hugging Face. Compare models, licenses, context windows, benchmarks, and free access in 2026.https://freeainews.com/open-source/hugging-face-3d-printed-humanoid-open-source/Sat, 22 Aug 2026 00:26:41 -0400https://freeainews.com/open-source/hugging-face-3d-printed-humanoid-open-source/Hugging Face launched a $2,500 open-source humanoid robot. It includes Apache 2.0 software, CERN OHL hardware, and 7B onboard models.https://freeainews.com/open-source/hidream-o1-image-open-source-2026/Sat, 22 Aug 2026 00:23:35 -0400https://freeainews.com/open-source/hidream-o1-image-open-source-2026/Lightricks releases HiDream-O1-Image, an open 8B image model that beats FLUX.2 on prompt fidelity and text rendering. Free Apache 2.0 weights.https://freeainews.com/open-source/hidream-o1-image-dev-2604-new-open-source-text-to-image-ai/Sat, 22 Aug 2026 00:19:56 -0400https://freeainews.com/open-source/hidream-o1-image-dev-2604-new-open-source-text-to-image-ai/Lightricks releases HiDream-O1-Image-Dev-2604 under Apache 2.0 with 512-token prompts, 8B weights, and benchmark wins over paid APIs.https://freeainews.com/open-source/hermes-agent-nous-research-open-source-2026/Sat, 22 Aug 2026 00:16:17 -0400https://freeainews.com/open-source/hermes-agent-nous-research-open-source-2026/NousResearch released Hermes Agent, a free open-source AI agent with 8B and 70B models, 131k context, and Apache 2.0. See benchmarks and setup.https://freeainews.com/open-source/headroom-llm-token-compression-open-source-2026/Sat, 22 Aug 2026 00:12:57 -0400https://freeainews.com/open-source/headroom-llm-token-compression-open-source-2026/Headroom is a new open source tool that claims up to 95% LLM token cost reduction. See how it works, who released it, and what it means for developers.https://freeainews.com/open-source/exaone-4-5-lg-ai-research-open-weight-vision-language-model/Sat, 22 Aug 2026 00:07:48 -0400https://freeainews.com/open-source/exaone-4-5-lg-ai-research-open-weight-vision-language-model/LG AI Research releases EXAONE 4.5, a 32B open-weight vision language model with 128K context, strong document reasoning, and free commercial use.https://freeainews.com/open-source/cohere-command-a-plus-open-source-2026/Sat, 22 Aug 2026 00:05:02 -0400https://freeainews.com/open-source/cohere-command-a-plus-open-source-2026/Cohere Command A+ is a free 218B MoE model that runs on two H100s. Explore its license, benchmarks, and how it compares to DeepSeek V4 and others.https://freeainews.com/open-source/glm-5-open-source-mit-2026/Fri, 21 Aug 2026 22:37:32 -0400https://freeainews.com/open-source/glm-5-open-source-mit-2026/Zhipu AI released GLM-5, an MIT licensed open source AI model with 1.4T parameters and 256K context, beating many paid models.https://freeainews.com/open-source/free-ai-data-analysis-open-source-tools-2026/Fri, 21 Aug 2026 22:30:46 -0400https://freeainews.com/open-source/free-ai-data-analysis-open-source-tools-2026/Compare the best free open-source AI data analysis tools: features, benchmarks, license, and how to run them with no API costs.https://freeainews.com/open-source/deepseek-v4-open-source-2026/Fri, 21 Aug 2026 22:23:59 -0400https://freeainews.com/open-source/deepseek-v4-open-source-2026/DeepSeek V4 is a 1.6T-parameter open model with 256K context, free weights, and GPT-5.5-rival benchmarks. We compare cost, license, and self-hosting.https://freeainews.com/open-source/best-open-source-llm-models-2026-coding-local-agentic-ai-benchmarks-and-license/Fri, 21 Aug 2026 22:15:38 -0400https://freeainews.com/open-source/best-open-source-llm-models-2026-coding-local-agentic-ai-benchmarks-and-license/Compare top open-weight LLMs for coding, local use, and agentic work in 2026. Benchmarks, licenses, and hardware requirements.