# Changelog

> Review verified ScaleMax changes to API behavior, models, clients, billing, documentation, and operational features.

Public, customer-facing changes to ScaleMax, newest first. Internal refactors, dependency bumps, and infrastructure work aren't listed here.

## September 2026

- **GPT-5.5, GPT-5.4, and GPT-5.4 mini retired.** From September 30, `gpt-5.5`, `gpt-5.4`, and `gpt-5.4-mini` are no longer served. Switch GPT-5.5 and GPT-5.4 mini to `gpt-5.6-luna` (same Luna card) and GPT-5.4 to `gpt-5.6-terra`.
- **Local Models family.** One token pack now covers the live Local Models roster (52 public ids, grouped by maker on the [models](https://docs.scalemax.pro/docs/models) page), including DeepSeek, Gemini, Grok, GLM, MiniMax, Kimi, Hunyuan, MiMo, Mistral, Nemotron, Qwen, Muse, StepFun, Meituan, and GPT OSS.
- **Catalog withdrawals.** Public ids whose upstreams no longer serve them were removed from advertised model lists. Fetch the authenticated catalog for the live set.
- **GPT-5.5 and GPT-5.4 mini.** Both settle on the Luna customer card ($0.522 / $3.132 per 1M, cached $0.0522, long-context $1.044 / $4.698 above 272K).

## August 2026

- **Gemini, Grok, and Qwen models.**Google Gemini (Flash, Flash Lite, and Pro Preview lines), xAI's Grok 4.5, and Alibaba's Qwen 3.6 Plus and Qwen 3.7 Plus are live in the authenticated catalog, with published customer rates and their own provider-scoped token packs.

## July 2026

- **Documentation search.** Press `/` or `Cmd/Ctrl K` anywhere in the docs to jump to a page. The command palette matches page and section titles as you type.
- **Guided setup wizard.** Sign in once with your key at `dashboard.scalemax.pro/setup` to see your matching credit or token base URL, provider scope, and live key-scoped model list, with links straight to Web Chat, Studio, and client-specific setup.
- **Redeem codes.** Apply a redeem code to add credit or a token pack to your key without going through checkout.

## June 2026

- **Studio.** Generate websites and apps from a prompt, preview them in the browser, and iterate — no local SDK required.
- **Web Chat.** Workspace-scoped conversations in the browser, billed to your active key. Pick a model per conversation and track spend as you go.
- **Usage checker.**Check a key's status and expiry, plus cumulative request and token counts, from the dashboard so you can reconcile usage against your credit balance or token budget.

## May 2026

- **Anthropic-compatible endpoint.** Native Anthropic Messages support alongside the OpenAI-compatible routes, so Claude clients can point at ScaleMax without a translation layer.
- **Faster Claude clients.** Native streaming, connection reuse, and disconnect cancellation for lower latency and cleaner request accounting.
- **Client setup guides.** Manual configuration pages for Codex, Claude Code, Cursor, Roo Code, Continue, TRAE SOLO, Claude Desktop, OpenCode, OpenClaw, Hermes, Cherry Studio, and direct API code — plus the `scalemax` CLI to configure them for you.

## April 2026

- **OpenAI-compatible gateway.** Chat completions, model listing, and streaming over the standard OpenAI protocol, with stable ScaleMax model names in front of upstream providers.
- **Two billing modes.** Pay from a USD credit balance or a fixed token budget. The gateway charges each key by the mode it was issued in, and prices are published per model.
- **Documentation site.** Launched `docs.scalemax.pro` with quick start, authentication, API reference, SDK examples, billing, rate limits, and troubleshooting.