Claude Opus 5 replaces Opus 4.8, and Kimi K3 arrives

Two hosted models landed at the end of July. Compare them on the models page.

Claude Opus 5

Opus 5 replaces Opus 4.8 in the hosted catalog at the same rates: $5 input, $25 output, $0.50 cache read, and $6.25 cache write per million tokens. Every stored reference to Opus 4.8, like a member's default model or a workflow's configured model, now resolves to Opus 5, so nothing needs reconfiguring.

• A 1,000,000-token context window.
• Reasoning effort from low to max, defaulting to medium.
• If Anthropic is overloaded, runs fall back to GPT-5.6 Sol, then Gemini 3.5 Flash.
• Opus 4.8 remains available through Token Tunnel for members whose Claude plan still serves it.

Kimi K3

Kimi K3 is Moonshot's open-weight model, available for chat, building, workflow runs, and as a utility model.

• A 1,048,576-token context window.
• Reasoning effort from low to max, defaulting to high.
• Image input.

Both models bill on actual cache reads and writes, like the rest of the catalog. Read why we think running open-source models yourself is a safe, cheap option.