传 Google 内部测试 Gemini 4 新检查点 Carbon，编码对标 Opus 5.5
爆料称 Google 员工在测比 Argon 更新的 Gemini 4 检查点 Carbon，测试者称编码手感像 Opus 5.5，若属实则 Gemini 可能同时兼顾多模态与编码。
Google might already have an Opus 5.5-level coding model... and it isn't Gemini 4 Argon. Google employees are testing a newer Gemini 4 checkpoint called Carbon. one tester said it “feels like Opus 5.5” for coding. another: “Carbon is really good!” Opus 5.5 is still my favorite coding model ever... but Gemini has never really had Claude's or ChatGPT's reputation for coding. Gemini has always been heavily trained as an everyday model. > multimodal. > huge context. > research. > Google's entire ecosystem. > science now imagine THAT model also being Opus-level at coding. you wouldn't have to pick: Claude for serious coding. Gemini for basically everything else. you could just use Gemini for both. and we haven't even properly gotten Argon yet. Carbon is already newer. if those early coding reports are real... Google might finally have the model that does everything. Will Google seriously make a comeback?