Axolotl v0.16 vs Honker

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.

Honker is a Rust-built SQLite extension that brings Postgres-style NOTIFY/LISTEN semantics to SQLite without any external broker. It adds cross-process notifications, durable pub/sub channels, task queues with retries and priority, and crontab-style scheduling — all living inside your existing SQLite file. Single-digit millisecond delivery via WAL-file watching instead of polling. The core trick: rather than polling the database on an interval, Honker watches SQLite's Write-Ahead Log (WAL) file with stat(2) calls. When a write lands, listeners wake up immediately. This gives push semantics without Redis, RabbitMQ, or any additional infrastructure. Business logic writes and task enqueues are atomic because they're in the same database. Honker ships as a loadable SQLite extension plus language packages for Python, Node.js, Rust, Go, Ruby, Bun, Elixir, and C++. It's experimental and the API may change, but it's addressing a real pain point: SQLite projects that outgrow simple reads/writes inevitably reach for external messaging, and Honker defers that moment significantly.