Search Ctrl+K
Change language Switch ThemeSign In
Curated Daily BriefWeekly PicksTopicsWorld Cup Special SettingsHelp CenterCollapse
Narrow Mode
Anthropic Releases Claude Sonnet 5: Agent Capabilities Nearly on Par with Opus, Pricing Lowest Among Peers
Anthropic Releases Claude Sonnet 5: Agent Capabilities Nearly on Par with Opus, Pricing Lowest Among Peers
 宝玉@dotey
Anthropic 今天发布 Claude Sonnet 5，替代 Sonnet 4.6 成为免费版和 Pro 版的默认模型。Anthropic 的定位很明确：Agent 能力接近自家最贵的 Opus 4.8，API 价格只有后者的 40%。
Sonnet 系列是开发者用量最大的一档。但过去几个月，AI Agent 能力（让模型自主规划、调用工具完成多步骤任务）的主要进步集中在更贵的 Opus 系列，两者差距越来越明显。Sonnet 5 把差距缩了回来。在 Agent 编程基准上，Sonnet 5 得分 63.2%，Sonnet 4.6 是 58.1%，Opus 4.8 是 69.2%。在知识工作基准上，Sonnet 5 甚至略微超过了 Opus 4.8。
早期测试者的反馈比较一致：以前 Sonnet 做到一半会停的复杂任务，现在能跑完，还会主动检查自己的输出。Zapier 的工程师说，让 Sonnet 5 连续执行“更新 Salesforce 账户等级，再给企业客户发公告邮件”，模型一口气做完了，“以前会卡在半路”。
API 定价分两阶段：8 月 31 日前的推广价是输入 2 美元/百万 Token、输出 10 美元/百万 Token，之后涨到 3 美元和 15 美元。据 TechCrunch 报道，这个价格低于 OpenAI 的 GPT-5.5 和 Google 的 Gemini 3.1 Pro，但仍高于 Gemini 3.5 Flash。
有个容易忽略的细节：Sonnet 5 换了新的分词器，同样的文本可能消耗 1.0 到 1.35 倍的 Token。Anthropic 说推广期的定价已经把这个涨幅对冲掉了，过渡期总成本大致不变。但推广价结束后，实际花费会比官方标价的涨幅更大。
安全方面，Sonnet 5 的幻觉率和迎合倾向低于前代，Agent 场景下抵御提示注入和恶意请求的能力更强。因为网络安全能力有所提升，模型默认开启了实时安全防护（和 Opus 4.7、4.8 相同的机制）。
Sonnet 5 今天起在 Claude 所有套餐、Claude Code 和 API 上可用，模型代号 claude-sonnet-5。Show More
#### Claude
@claudeai · 6h ago
Introducing Claude Sonnet 5, our most agentic Sonnet yet.
It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models.Show More
00:11
1,597
3,206
31.4K
3.6M
Jun 30, 2026, 6:33 PM View on X
15 Replies
6 Retweets
60 Likes
23.9K Views  宝玉 @dotey
Follow
One Sentence Summary
Anthropic releases Claude Sonnet 5, positioned as the most agentic Sonnet model, approaching Opus 4.8 on coding benchmarks, with API pricing lower than major competitors.
Summary
Anthropic today released Claude Sonnet 5, replacing Sonnet 4.6 as the default model for the free and Pro tiers. Its agentic capabilities are close to the most expensive Opus 4.8, while the API price is only 40% of the latter. On the agentic coding benchmark, Sonnet 5 scored 63.2%, compared to Opus 4.8's 69.2%, narrowing the gap. Early testers report that complex tasks that previously stopped midway can now complete, with the model proactively checking its own output. API pricing has two phases: introductory pricing of $2/M input tokens and $10/M output tokens until August 31, then rising to $3/M and $15/M. Sonnet 5 uses a new tokenizer, which may consume 1.0 to 1.35 times the tokens for the same text; the introductory pricing offsets this, but after it ends, actual costs may increase more than the official price hike suggests. In terms of safety, hallucination rate and sycophancy tendency have decreased, and resistance to prompt injection in agentic scenarios has improved. Sonnet 5 is available today on all Claude plans, Claude Code, and the API.
AI Screening
89
Influence Score 18
Published Today
Language
Chinese
Tags
Claude Sonnet 5
Anthropic
AI Model
Agent Capabilities
API Pricing
Make your daily reading actually fit you.A daily brief built from the sources you follow. Get started free HomeDiscoverWorld CupSettings