This week, large models may be released in clusters, with GPT-6 Sol, Opus 5.5, and Grok 4.7 all queuing up.

Beating AI News Flash: This week may see a wave of intensive large model updates, with at least 7 models showing release or testing signals.

Tier 1: GPT-6 Sol, Claude Opus 5.5. OpenAI has not yet confirmed GPT-6 Sol, but Sam Altman previously said the "major thing" originally scheduled for release last week was postponed to this week; the community has recently seen continuous leaks about GPT-6 Sol testing and gray-scale routing. On Anthropic's side, blogger Lyra claims Opus 5.5 is being tested as claude-wafer-eap and is planned for release on Tuesday. Reuters also reported last week that Anthropic is considering launching a new model soon, but did not confirm the name Opus 5.5.

Tier 2: Grok 4.7, Qwen4, Kimi K3.1. Musk originally said Grok 4.7 would launch around September 12, then later said it would take "a few more days," and it has still not been released. Qwen has explicitly stated that Qwen3.8-Flash-Next is an early preview of the Qwen4 architecture, and September 22 to 24 coincides with the Yunqi Conference, though officials have not previewed that Qwen4 will be released at the conference. Kimi's recent π hints have already teased K3.1, but there is currently no release date, model card, or API.

Tier 3: Gemini 4 Pro, DeepSeek V4.1 Pro. Gemini 4 Pro was recently reported to be testing in Arena under codenames such as Argon, but Google has so far only confirmed that Gemini 4 is in training and has not announced a release date. DeepSeek has already directly mentioned the future V4.1 Pro in its official documentation, but has likewise not given a launch time.