GPT-6 Astra and Claude Fable 5.1 available on hosted keys

GPT-6 Astra and Claude Fable 5.1 are now hosted models: any workspace can run them on General Input's provider keys without connecting a subscription. Until now both were reachable only through Token Tunnel. Compare them on the models page.

What changed

• GPT-6 Astra: a 1,050,000-token context, at $10 per million input tokens and $50 output, with cache reads at $1. Above 272,000 input tokens, input bills at 2x and output at 1.5x.
• Claude Fable 5.1: a 1,000,000-token context and up to 128,000 output tokens, at $10 input and $50 output, with cache reads at $0.25.
• Same-model failover. Both are also served through Vercel AI Gateway and OpenRouter, so an overload moves the same model to another route before any fallback.
• Long-context pricing is now a catalog field: an input threshold plus multipliers. Each request in a long run is priced against its own context size.

Also shipped

• GLM 5.3 replaces GLM 5.2, and GLM 5.3 Flash joins with a 1M-token context.
• DeepSeek V4.1 Flash replaces V4 Flash, adds vision, and costs half as much: $0.22 input and $0.66 output per million tokens.
• Token Tunnel now delivers instructions to Claude and ChatGPT subscriptions as a real system prompt.