AITOPAI热报
Models

Google 发布 Gemini 4 Carbon，主打长程任务与多步推理

Paul Couvert@itsPaulAi
Post
View the post
Why we picked it

Google 一个月内连发两个模型，这次 Carbon 主打长程任务和多步推理。有人用它在一个 HTML 文件里纯代码生成了整段视频，不用任何剪辑工具。如果你在做智能体或者多步自动化，这个能力点值得关注。

Google 官方宣布 Gemini 4 Carbon 现已可用，定位为其最强模型，重点提升深度推理、长程任务处理与多步推理能力。官方表述是：如果 Argon 适合复杂任务，Carbon 则用于攻克最困难的问题。Argon 发布至今不到一个月，Carbon 紧随其后推出，Google 的模型迭代节奏明显加快。

Image · Paul Couvert
Body

从实际体验看，博主 Paul（itsPaulAi）提前试用数天，展示的例子中，Carbon 在单个 HTML 文件里用代码完整生成了一段视频，全程没有借助外部工具或剪辑流程。这说明该模型在代码生成与多步任务串联上有较强的一致性，能在一整个文件内维持长链路逻辑不崩。

对开发者而言，长程任务和多步推理正是智能体工作流最依赖的能力，Carbon 与 Argon 的分层定位（常规复杂任务 vs 最难问题）给了使用者明确的选型参考。与 OpenAI、Anthropic 的高端模型相比，Google 这次强调的是任务纵深而非单点跑分。

Original · Paul Couvert

Google has just released Gemini 4 Carbon This model produced the video below 100% using code... in a single HTML file. I've had the chance to use it for the past few days. And its capabilities are REALLY impressive. Keep in mind that Argon was launched less than a month ago! Feel the RSI. Thanks to Google for the early access!

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Gemini 4 Carbon, Google, 推理模型, 智能体.

Source attribution: curated from Paul Couvert (@itsPaulAi).

#Gemini 4 Carbon#Google#推理模型#智能体

Keep reading

Recent stories with the same tag or section
ModelsUS vs ChinaMulti-source00:05
Chinese contentGoogle 公开发布 Gemini 4 Carbon，主打智能体“蜂群”功能

Google 把只给安全防御者用两周的 Gemini 4 Carbon 开放给所有人了，多智能体蜂群协作是这次最大的看点，可以直接拿来和 Claude、OpenAI 的旗舰模型对比跑分，做智能体应用的开发者值得上手试试。

#Gemini 4 Carbon#Google#Anthropic#OpenAI
Event
Official accounttheverge@Robert HartSource
ModelsUS vs ChinaMulti-source86°03:50
Chinese contentAnthropic 发布 Claude Fable 5.2，登顶 Artificial Analysis 智能指数

Anthropic 的新旗舰模型，科学推理和编程基准刷了一堆纪录，比如 HLE-Diamond 拿到 80%。有意思的是调低 effort 档反而更划算，输出 token 用得更少还能赢 Opus 5.5。做科学计算、代码开发的团队值得看下这笔成本账。

#Claude Fable 5.2#Anthropic#Gemini 4 Carbon#推理模型
Event
Official accountArtificial Analysis@ArtificialAnlysSource
ModelsBuildersMulti-sourceSelected02:23
Agent Arena data shows 66% cost gap in hour-long agent tasks

For hour-long agent workloads, the cheapest model is not automatically the right one: Claude Opus 5.5 scores higher but costs twice as much per hour as Gemini 4 Carbon. Teams running long-horizon agents can use these per-hour cost and time figures to match model choice to budget and quality needs. The 55-minute spread among frontier models also suggests task duration, not just accuracy, should enter model selection criteria.

#Claude Opus 5.5#Gemini 4 Carbon#GPT-6 Astra#Kimi K3.1 Pro
Event
lmarena.ai@lmarena_aiSource