Axolotl v0.16 vs Tendril

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.

Tendril is an open-source desktop agent built on a radically minimal architecture: instead of giving an AI model dozens of pre-built tools, it gives the model exactly three — search capabilities, register capabilities, and execute code. When you ask it to do something it can't yet do, it writes the tool, registers it, and runs it. The next time you ask for something similar, the tool already exists. Built with Tauri, React, and Node.js on the frontend, and AWS Bedrock (Claude) for inference, Tendril runs code in sandboxed Deno environments for safety. The capability registry grows organically across sessions, meaning the agent becomes measurably more capable the longer you use it — without any retraining or fine-tuning. The "too many tools" problem is a real issue in production agents: large tool lists degrade model reasoning and increase hallucination rates. Tendril's inversion of this pattern — grow tools from need, not configuration — is a genuine architectural contribution. It's MIT licensed and free to use, though AWS Bedrock access for Claude adds ongoing inference costs.