Axolotl v0.16 vs tldr MCP Gateway

Axolotl is the go-to open-source fine-tuning framework for the local LLM community, and v0.16 is its most significant performance release to date. The headline numbers are striking: 15x faster training for Mixture-of-Experts (MoE) models with LoRA adapters, 40x reduction in memory usage for the same configurations, and 58% faster GRPO async training — the algorithm behind many of the recent reasoning model breakthroughs. Day-0 support for Google Gemma 4 shipped simultaneously with the model release. The MoE+LoRA improvements are especially timely. As sparse mixture-of-experts models like Gemma 4, Mistral, and Qwen3.6-Plus dominate the model landscape, fine-tuning them has been disproportionately expensive. Axolotl v0.16 makes it practical to fine-tune these architectures on a single consumer GPU — previously a multi-GPU or cloud-required task. The GRPO improvements also make reinforcement learning from human feedback (RLHF) workflows dramatically faster for small teams. For the indie fine-tuning community — researchers, small companies, and hobbyists building specialized models — this release removes a major cost barrier. Combined with the simultaneous Gemma 4 support, v0.16 positions Axolotl as the fastest path from a new model release to a fine-tuned, production-ready custom variant.

tldr is a local proxy that sits between your AI coding harness and upstream MCP servers, solving one of the most underappreciated problems in agentic workflows: context bloat from tool schema proliferation. When you connect GitHub MCP, filesystem MCP, and a few others, you can easily be sending 24,000+ tokens of tool schemas to the model before any work begins. Instead of passing all those schemas directly, tldr exposes exactly five wrapper tools to the model: search_tools, execute_plan, call_raw, inspect_tool, and get_result. The model learns which underlying tools exist on-demand through search_tools, then calls them through the proxy. GitHub MCP's 24,473-token schema surface compresses to 3,482 tokens — an 86% reduction. Output responses are further compressed through field stripping, a 4,096-token cap, and a 64KB byte limit. This is a genuinely practical solution for power users running multi-MCP setups who've noticed degraded performance as their tool count grows. The tradeoff is one extra hop of indirection, but the token savings pay for themselves in improved model attention and lower API costs.