Changelog / release

1.1.0 — Inkling 2 Mini and a supervised server

Released 2026-11-21 · full changelog · GitHub releases

Upgrade: pip install -U rapid-mlx  ·  brew upgrade rapid-mlx  ·  or grab the desktop app.

Rapid-MLX 1.1.0 adds native Inkling 2 Mini, keeps the model runner supervised, and brings status, recent API calls and agent jobs into one local control surface.

  • Inkling 2 Mini. Serve inkling-2-mini-4bit with the bundled native implementation, without executing checkpoint code. This release supports text only; reasoning is separate from the answer and tool calls use OpenAI tool_calls in streaming and non-streaming responses. Model guide. (#5907, #5908)
  • Supervised serve. A small parent process restarts a crashed runner on the same port. Ctrl-C stops the runner and frees the port. rapid-mlx status and GET /rapid/v1/status expose runner state and restart information. Control commands. (#5897)
  • Recent API calls. rapid-mlx log and GET /rapid/v1/log show up to 1,000 calls in runner memory: time, method, path, status, duration and client label. Bodies, query strings and other headers are never recorded. (#5900)
  • Agent jobs. rapid-mlx jobs and GET /rapid/v1/jobs list agent runs; rapid-mlx jobs cancel <id> cancels one. (#5906)
  • Background priority. Send X-Rapid-Priority: background for batch work. Waiting interactive requests are admitted first; after 30 seconds, a waiting background request is treated as interactive. Running work is not preempted. API details. (#5901)
  • Continue an assistant message. Chat completions accept continue_final_message for a partial final assistant message. Set add_generation_prompt: false to continue that message instead of opening a new turn. Request example. (#5898)
  • Telemetry categories. Model download failures use fixed failure subtypes, and document inputs are counted separately from image inputs. No free text is added. The complete disclosure lists the published values. (#5903)
  • Getting started. The README opens with a three-step path for coding agents and SDKs. (#5904)
  • Upgrade to 1.1.0. Released 2026-11-21. Run rapid-mlx upgrade, brew upgrade rapid-mlx or pip install -U rapid-mlx==1.1.0. Full release notes · Download Desktop 1.1.0.