Can a local LLM replace Claude Code? The benchmark everyone quotes is the wrong one.

说明与用途

本地LLM可以生成正确代码，但在可靠地发出工具调用、保持长周期一致性、处理大上下文以及遵循指令方面表现不佳。2026年，Mixture‑of‑Experts模型和本地API支持使得在消费级GPU上运行大型模型成为可能，但选择合适的模型系列并配置上下文窗口对实际编辑器自动化至关重要。

问题

Local LLMs can generate correct code but fail to reliably act as agents that emit valid tool calls, maintain long‑horizon consistency, handle large context, and follow instructions, making them unsuitable for real‑world editor automation.

结果

A single 24GB GPU running Qwen 3.6 27B provides the sweet spot for reliable local coding agents, offering ~77% SWE‑bench score and robust tool‑calling, while smaller GPUs can use GLM‑4.7‑Flash or Gemma 4 for single‑file help.

工作流

• Identify GPU VRAM and desired agentic capabilities
• Choose a model family and size (e.g., Qwen 3.6 27B for 24GB GPU)
• Pull the model with Ollama (e.g., ollama pull qwen3-coder:30b)
• Set environment variables to point Claude Code to localhost (export ANTHROPIC_BASE_URL="http://localhost:11434", export ANTHROPIC_AUTH_TOKEN="local-dummy")
• Run Claude Code (claude) or Codex CLI with --oss flag
• Configure context window to at least 32K or 64K
• Verify tool‑call reliability and consistency in practice
• Adjust model or harness if tool‑call failures occur

证据

SWE‑bench measures "given a bug, produce a correct patch." That's a *writing* task.

来源

Tool‑calling is a separate trained skill from writing code.

来源

The curve climbs steeply up to ~24GB and then flattens — the 480B coder even dips below the 27B on Verified while costing 3× the memory.

来源