https://localclaw.io/new.html Recently released open-weight AI models verified for local use in the LocalClaw catalogue. en Wed, 01 Jul 2026 12:00:00 GMT 1440 https://localclaw.io/models/hy3 https://localclaw.io/models/hy3 Wed, 01 Jul 2026 12:00:00 GMT hy3 Tencent Hy Team MoE model with 256K context, strong agent/productivity benchmarks and Apache 2.0 licensing. Practical only for very large local workstations via IQ1_M GGUF. 295B (21B active) · 128 GB minimum RAM · IQ1_M · hy3. https://localclaw.io/models/bonsai-27b https://localclaw.io/models/bonsai-27b Wed, 01 Jul 2026 12:00:00 GMT bonsai PrismML low-bit model derived from Qwen 3.6 27B. Official Apache 2.0 ternary (7.2GB deployed) and 1-bit (3.9GB) builds retain multimodal, reasoning and agentic capabilities through custom GGUF and MLX runtimes. 27.3B (ternary / 1-bit) · 16 GB minimum RAM · Ternary Q2_0_g128 · bonsai. https://localclaw.io/models/laguna-xs-2.1 https://localclaw.io/models/laguna-xs-2.1 Wed, 01 Jul 2026 12:00:00 GMT laguna Poolside agentic coding MoE with 262K context, 33B total / 3B active parameters, OpenMDW-1.1 licensing and official Q4_K_M GGUF plus Ollama availability for 36GB-class local machines. 33B (3B active, MoE) · 36 GB minimum RAM · Q4_K_M · laguna. https://localclaw.io/models/lfm2.5-8b-a1b https://localclaw.io/models/lfm2.5-8b-a1b Mon, 01 Jun 2026 12:00:00 GMT lfm Liquid AI hybrid model built for on-device assistants. 8.3B total / 1.5B active, 128K context, tool use, GGUF, ONNX, MLX, llama.cpp and LM Studio support. Open-weight under LFM 1.0. 8.3B (1.5B active) · 8 GB minimum RAM · Q4_K_M · lfm. https://localclaw.io/models/qwen-agentworld-35b-a3b https://localclaw.io/models/qwen-agentworld-35b-a3b Mon, 01 Jun 2026 12:00:00 GMT qwen Official Qwen language world model for simulating agent environments across terminal, web, OS, Android, search, SWE and tool-calling domains. Apache 2.0 with active GGUF and MLX quantizations. 35B (3B active, MoE) · 32 GB minimum RAM · Q4_K_M · qwen. https://localclaw.io/models/north-mini-code-1.0 https://localclaw.io/models/north-mini-code-1.0 Mon, 01 Jun 2026 12:00:00 GMT cohere Cohere Labs Apache 2.0 coding and agent model. 30B total / 3B active MoE, 256K context, terminal-task training and mature GGUF quantizations for local workstation use. 30B (3B active, MoE) · 32 GB minimum RAM · Q4_K_M · cohere. https://localclaw.io/models/agents-a1 https://localclaw.io/models/agents-a1 Mon, 01 Jun 2026 12:00:00 GMT internscience InternScience Apache 2.0 agentic VLM. 35B-A3B MoE, 262K context, strong long-horizon search/tool-use benchmarks and official Q4_K_M GGUF artifacts for local workstations. 35B (3B active, MoE) · 32 GB minimum RAM · Q4_K_M · internscience. https://localclaw.io/models/diffusiongemma-26b-a4b-it https://localclaw.io/models/diffusiongemma-26b-a4b-it Mon, 01 Jun 2026 12:00:00 GMT gemma Official Google Apache 2.0 diffusion-language Gemma model with image-text chat support. Strong local relevance thanks to active Unsloth GGUF quantizations for workstation-class machines. 26B (4B active, diffusion MoE) · 32 GB minimum RAM · Q4_K_M · gemma. https://localclaw.io/models/gemma4-12b https://localclaw.io/models/gemma4-12b Mon, 01 Jun 2026 12:00:00 GMT gemma Google DeepMind 12B unified multimodal model. Text, image, audio and video inputs, 256K context, Apache 2.0, and a strong local sweet spot for 16-32 GB machines. 12B · 16 GB minimum RAM · Q4_K_M · gemma. https://localclaw.io/models/glm-5.2 https://localclaw.io/models/glm-5.2 Mon, 01 Jun 2026 12:00:00 GMT glm Z.ai flagship open model for long-horizon coding, reasoning and agentic work. 744B total, 40B active, 1M-token context, MIT license. Unsloth Dynamic GGUF makes it technically local, but it needs workstation/server-class memory: ~245GB total memory for 2-bit and 372GB+ for 4-bit. 744B (40B active) · 256 GB minimum RAM · UD-IQ2_M · glm. https://localclaw.io/models/ornith-1.0-9b-gguf https://localclaw.io/models/ornith-1.0-9b-gguf Mon, 01 Jun 2026 12:00:00 GMT deepreinforce Compact Ornith 1.0 GGUF variant from DeepReinforce for agentic coding experiments on consumer hardware. MIT licensed and much more practical than the frontier 397B release. 9B · 8 GB minimum RAM · Q4_K_M · deepreinforce. https://localclaw.io/models/ornith-1.0-35b-gguf https://localclaw.io/models/ornith-1.0-35b-gguf Mon, 01 Jun 2026 12:00:00 GMT deepreinforce DeepReinforce Ornith 1.0 mid-size GGUF release for agentic coding. The Q4_K_M build is listed around 21.2GB, making it a realistic 32GB+ local model compared with the 397B server-grade version. 35B MoE · 32 GB minimum RAM · Q4_K_M · deepreinforce. https://localclaw.io/models/minicpm5-1b https://localclaw.io/models/minicpm5-1b Fri, 01 May 2026 12:00:00 GMT minicpm OpenBMB compact on-device LLM with Apache 2.0 licensing, 128K context, tool-calling focus and official GGUF plus MLX artifacts for laptops and edge devices. 1B · 4 GB minimum RAM · Q4_K_M · minicpm. https://localclaw.io/models/deepseek-v4-flash https://localclaw.io/models/deepseek-v4-flash Fri, 01 May 2026 12:00:00 GMT deepseek-flash Efficient DeepSeek V4 variant: 284B total, 13B active, 1M-token context. Flash-Max can approach Pro reasoning with larger thinking budget. MIT licensed. 284B (13B active) · 256 GB minimum RAM · FP4/FP8 · deepseek-flash. https://localclaw.io/models/dante-mosaic-3.5b https://localclaw.io/models/dante-mosaic-3.5b Fri, 01 May 2026 12:00:00 GMT dante OdaxAI compact dense model based on SmolLM3-3B and distilled from Kimi K2. Strong small-model benchmark profile: GSM8K 74.45, HellaSwag 76.73 and MBPP 42.6. Apache 2.0, BF16 weights, practical for local Transformers/vLLM use. 3.08B · 8 GB minimum RAM · BF16 · dante. https://localclaw.io/models/granite4.1-8b https://localclaw.io/models/granite4.1-8b Wed, 01 Apr 2026 12:00:00 GMT granite IBM Granite 4.1 long-context instruct model. Apache 2.0, 131K context, tool calling, RAG, code tasks, multilingual dialog and business assistant workflows on normal 8-16 GB machines. 8B · 8 GB minimum RAM · Q4_K_M · granite. https://localclaw.io/models/qwen3.6-6.7b https://localclaw.io/models/qwen3.6-6.7b Wed, 01 Apr 2026 12:00:00 GMT qwen Alibaba's hybrid-thinking micro-flagship. Toggles between instant answers and deep chain-of-thought reasoning on demand. 128K context, 29 languages, outperforms Qwen3-8B on reasoning benchmarks. Apache 2.0. 6.7B · 8 GB minimum RAM · Q4_K_M · qwen. https://localclaw.io/models/qwen3.6-27b https://localclaw.io/models/qwen3.6-27b Wed, 01 Apr 2026 12:00:00 GMT qwen Qwen 3.6 flagship dense model. Hybrid thinking mode with /think toggle for deep chain-of-thought reasoning. 128K context, 29+ languages. Significantly outperforms Qwen3.5-27B on reasoning, coding & math. Apache 2.0. 27B · 32 GB minimum RAM · Q4_K_M · qwen. https://localclaw.io/models/qwen3.6-35b-a3b https://localclaw.io/models/qwen3.6-35b-a3b Wed, 01 Apr 2026 12:00:00 GMT qwen Qwen Team open-weight MoE for agentic coding and multimodal work. 35B total / 3B active, 262K native context, Apache 2.0, and strong GGUF availability through Unsloth and LM Studio-compatible artifacts. 35B (3B active, MoE) · 32 GB minimum RAM · Q4_K_M · qwen. https://localclaw.io/models/ling-2.6-flash https://localclaw.io/models/ling-2.6-flash Wed, 01 Apr 2026 12:00:00 GMT ling InclusionAI's MIT-licensed instruct MoE optimized for fast agent workloads. 104B total parameters, only 7.4B active, hybrid linear attention, 262K context and strong tool-use / multi-step execution with high token efficiency. 104B (7.4B active) · 80 GB minimum RAM · Q4_K_M · ling.