Best local AI models for coding, repo work, debugging and software engineering. Compare RAM, quality, coding scores and LM Studio setup. Ranked from the LocalClaw model database with RAM requirements, quantization and links to static model pages.
For coding, start with Kimi K2 Instruct (1T MoE) if your hardware fits it. If not, choose the highest-ranked model that fits your RAM tier and preferred quantization.
Moonshot AI trillion-parameter MoE flagship. 32B active params per token with 384 experts. Matches or beats GPT-4 Turbo on MMLU, GSM8K, HumanEval. Agentic & tool-use specialist. Server-grade only. Modified MIT.
MiniMax native multimodal MoE with 1M context and MiniMax Sparse Attention. Around 428B parameters with 23B active. Built for long-context coding, cowork and agentic workflows, with local deployment via SGLang, vLLM or Transformers. Server-grade only.
Moonshot AI coding-focused agentic Kimi built on K2.6. 1T MoE with 32B active parameters, 256K context, MoonViT vision encoder and stronger long-horizon coding while reducing thinking-token usage by roughly 30% vs K2.6. Modified MIT. Server-grade only.
Flagship open-source Qwen 3.5. Only 17B active params despite 397B total — world-class quality at MoE efficiency. Matches GPT-4o on major benchmarks. Requires multi-GPU or server-grade hardware. Apache 2.0.
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.
Moonshot AI K2 with extended reasoning mode. Chain-of-thought traces before final answer. Top-5 on GPQA, AIME, SWE-bench. Requires datacenter-grade hardware or distributed inference. Modified MIT.
DeepSeek frontier MoE with 1M-token context, hybrid compressed attention and top-tier coding/reasoning. MIT licensed. Datacenter-grade only.
Z.ai next-generation flagship for agentic engineering. Stronger coding, long-horizon tool use, SWE-Bench Pro, Terminal-Bench and repo generation. MIT licensed.
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.
Experimental V3.2 with DeepSeek Sparse Attention (DSA) — halves inference cost vs V3.1 on long context while keeping quality. 128K context, improved coding & tool-use. MIT licensed. Server-grade.
Zhipu AI flagship — full GLM 4.6. 200K context, strong tool-calling & agentic workflows. Competes with Claude 3.5 Sonnet on reasoning and code. MIT licensed. Server-grade hardware.
Updated flagship DeepSeek R1 with improved reasoning chains and fewer hallucinations. Major upgrade to chain-of-thought quality. MIT licensed. Server-grade only.
Near GPT-4 intelligence locally. Thinking mode demolishes hard problems. The local AI dream.
Moonshot AI's agentic flagship. 1T total MoE parameters with 32B active per forward pass. Unmatched long-context reasoning at 256K tokens. Designed for complex agentic tasks and tool use. Model License — check moonshotai.com for commercial terms.
Qwen flagship coding model. Designed for agentic coding with 256K context. Outperforms Claude 3.5 Sonnet on SWE-bench. Apache 2.0.
MiniMax MoE flagship with 10B active params and 4M-token long-context. Specialised for agentic coding and tool-use. Competitive with GPT-4 class models at a fraction of the inference cost. MIT licensed.
DeepSeek's massive MoE flagship. 37B active out of 671B total. Exceptional coding, reasoning and general capabilities. Ranks #6 on global usage leaderboards with 29B monthly tokens. MIT licensed.
Arcee AI's massive MoE open model. ~400B total parameters, 70B active per forward pass. Ranks near the top of global usage leaderboards. Exceptional versatility across reasoning, coding and chat. Free and open-source. Apache 2.0.
LocalClaw ranks models using their tags plus relative benchmark scores for speed, quality, coding and reasoning. The goal is a practical local setup recommendation, not a synthetic leaderboard.