Announcements

Model launches, retirements, routing improvements and service changes.

【GPT-6.1 Terra is live】You can now call gpt-6.1-terra. It accepts text and image input, supports tool calls, structured output and prompt caching, and has a context window of about 920K tokens. It works with the OpenAI, Anthropic and Gemini request formats, and two supply lines are available.

Search for GPT-6.1 Terra in Model Square to see live prices per line. Use the Balanced, Quality first or Cost first preset on your API key, or pick lines yourself.

2026-10-27T13:30:00+08:00

【17 new models this week, including Kling 3.0 and Kimi K3.1 Pro】We added 17 models this week: Kling 3.0, Jimeng and Vidu for video, six more for images, and Kimi K3.1 Pro and MiniMax M3.1 Flash for coding and chat. Two smaller changes: when an AI Studio job fails, its card now says whether you were charged and whether it was refunded, and you can now top up with Alipay.

2026-10-26T23:20:00+08:00

【Kimi K3.1 Pro is live】You can now call kimi-k3.1-pro through the OpenCode subscription line, with regular and streaming chat. It costs about ¥12.19 input, ¥62.62 output and ¥1.22 cache read per million tokens.

Search for Kimi K3.1 Pro in Model Square for live prices. Use the Balanced, Quality first or Cost first preset on your API key, or choose the OpenCode subscription line yourself.

2026-10-24T14:28:00+08:00

【Turn a story or a song into a video】AI Studio has two new things to try this week: Story to Video and Music Video. And if you use Claude Code, Codex or Gemini tools, some models used to reject your calls outright; that is fixed this week too.

2026-10-21T00:10:00+08:00

【Gemini 4 Argon is live】You can now call gemini-4-argon through supplier lines. It costs ¥1 input, ¥5 output and ¥0.05 cache read per million tokens, through /v1/chat/completions.

See Model Square for the lines, or choose one in your token's channel settings.

2026-10-14T01:14:02+08:00

【GPT audio models are live】gpt-audio and gpt-audio-1.5 are now available, with audio input and spoken output. Call them through /v1/chat/completions; text and audio are billed separately.

See Model Square for prices. Use a client that supports audio input and output.

2026-10-10T13:30:00+08:00

【Claude Haiku 5.5 is live】You can now call claude-haiku-5-5, from ¥0.1641 input and ¥0.8204 output per million tokens; when a request's input exceeds 100K tokens, from ¥0.15 and ¥0.75. Input length includes cached tokens, and each request is billed at its tier.

See Model Square for lines and cache prices, or choose a line in your token's channel settings.

2026-10-08T14:50:05+08:00

【Gemini image models are live】Gemini 3.1 Flash Image, Gemini 3 Pro Image and Gemini 3.1 Flash Lite Image are now available and generate images through the chat API, as gemini-3.1-flash-image, gemini-3-pro-image and gemini-3.1-flash-lite-image. A 1K image costs about ¥0.22, ¥0.44 and ¥0.11; most lines are about half of Google's official price.

Higher resolutions cost more per image. Google has announced that Gemini 3.1 Flash Image will be retired on October 29.

2026-10-07T21:00:00+08:00

【New models are live】Mistral Large 4, Qwen3.7 Flash and Muse Spark 1.2 / 1.3 are now available as mistral-large-4, qwen3.7-flash, muse-spark-1.2 and muse-spark-1.3. All are supplied through the Command Code subscription line at about 85% of the official price.

See line prices in Model Square, or choose the line in your token's channel settings.

2026-10-07T21:00:00+08:00