# Claude Opus 5.5, GPT-6 Sol and Luna, and Grok 4.7 arrive

> Opus 5.5 and GPT-6 replace Opus 5 and GPT-5.6 at lower per-token prices, Grok 4.7 replaces 4.6, and Auto tiers span four providers.

Sep 22, 2026. Topics: Models. https://www.generalinput.com/changelog/2026-09-22-opus-5-5-gpt-6-grok-4-7

We moved the catalog to the newest generation of each frontier model family, and saved selections followed on their own: a workflow, app, or specialist that names a retired model now runs on its successor, so nothing had to be re-saved. Subscription models through Token Tunnel follow the same upgrades. Compare everything on the [models page](/models).

### What changed

Prices are per million input and output tokens.

- **Claude Opus 5.5** replaces Opus 5 and 4.8 at $4 / $20, down from $5 / $25, with a 1M-token context.
- **GPT-6 Sol** replaces GPT-5.6 Sol at $2 / $10, down from $5 / $30.
- **GPT-6 Luna** replaces GPT-5.6 Luna at $0.10 / $0.50, down from $0.20 / $1.20. Both GPT-6 models have a 1.05M-token context.
- **Grok 4.7** replaces Grok 4.6 and 4.5 at the same $2 / $6, with a 500K context and a new xhigh effort level.
- **Claude Sonnet 5.5** replaced Sonnet 5 on September 28.
- **Default effort per model.** Every model used to default to medium; each now has its own default within its supported range.

### Auto

Each tier starts cheap on the cost-to-intelligence frontier and falls back across four providers (OpenAI, Anthropic, Google, xAI), so one provider's outage can't take a tier down. When one of your own subscription models is online, Auto uses it in place of the tier's cloud model.