[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fluxcvwzq361u":3},{"slug":4,"category":5,"publishedAt":6,"titleZh":7,"titleEn":8,"summaryZh":9,"summaryEn":10,"models":11,"hardwares":12,"bodyZh":13,"bodyEn":14,"sourceName":15,"sourceUrl":16,"updatedAt":17},"claude-opus-5-5","NEWS","2026-10-01T05:15:00.000Z","Anthropic 发布 Claude Opus 5.5：$4\u002F$20 定价，旗舰门槛再降","Anthropic Ships Claude Opus 5.5: $4\u002F$20 Pricing Lowers the Flagship Bar","Anthropic 发布 Claude 5.5 系列首款模型 Opus 5.5（claude-opus-5-5）：输入 $4\u002F输出 $20 每百万 token，较 Opus 5 降价 20%，1M 上下文，官方称多数工作负载达到 Fable 5.1 水平而价格不到一半；Sonnet 5.5 六天后跟进。","Anthropic has released Opus 5.5 (claude-opus-5-5), the first model of the Claude 5.5 family: $4\u002F$20 per million tokens — 20% below Opus 5 — with a 1M-token context window. Anthropic claims Fable 5.1-level results on most work at under half the price. Sonnet 5.5 followed six days later.",[],[],"## 发布\n\n2026 年 9 月 22 日，Anthropic 发布 **Claude Opus 5.5**（API 模型 ID：`claude-opus-5-5`），是 Claude 5.5 系列的首款模型，取代 Opus 5。**Claude Sonnet 5.5** 于 9 月 28 日跟进。值得注意的是，OpenAI 在 Anthropic 官宣后约一小时内发布了 GPT-6 Sol \u002F Luna——旗舰价格战的火药味很浓。\n\n## 定价（每百万 token）\n\n- 输入 **$4.00** \u002F 输出 **$20.00**——较 Opus 5（$5\u002F$25）降价 20%\n- 缓存读取 $0.20（较 Opus 5 低 60%）\n- Batch API 半价（$2.00\u002F$10.00）；Fast 模式 $8.00\u002F$40.00（输出速度最高 2.5×）\n- 官方称典型工作负载整体便宜约 40%（输出更快、单任务 token 更少）\n\n## 规格与能力\n\n- **1M token 上下文**，最大输出 128K（Batch API beta 可达 300K）\n- 自适应思考（adaptive thinking）默认开启且不可关闭，默认 effort 从 high 调整为 medium\n- 官方定位：多数工作上达到 **Claude Fable 5.1 水平**，而价格不到其一半（Fable 5.1 为 $10\u002F$50）\n- 知识截止 2026 年 6 月\n\n## 可用性与合规\n\n- Claude API、Amazon Bedrock、Google Cloud Vertex AI、Microsoft Azure\u002FFoundry，以及 Claude Pro\u002FMax\u002FTeam 订阅\n- 经 METR 与 Frontier Design 外部测试；内置水印以符合欧盟 AI 法案要求","## Launch\n\nOn September 22, 2026, Anthropic released **Claude Opus 5.5** (API model ID: `claude-opus-5-5`), the first model of the Claude 5.5 family, replacing Opus 5. **Claude Sonnet 5.5** followed on September 28. Notably, OpenAI announced GPT-6 Sol \u002F Luna within about an hour of Anthropic's announcement — the flagship price war is unmistakable.\n\n## Pricing (per million tokens)\n\n- Input **$4.00** \u002F output **$20.00** — 20% below Opus 5 ($5\u002F$25)\n- Cache reads $0.20 (60% lower than Opus 5)\n- Batch API at half price ($2.00\u002F$10.00); Fast mode $8.00\u002F$40.00 (up to 2.5× output speed)\n- Anthropic claims typical workloads run about 40% cheaper overall (faster output, fewer tokens per task)\n\n## Specs and Capabilities\n\n- **1M-token context window**, 128K max output (up to 300K via Batch API beta)\n- Adaptive thinking is always on and cannot be disabled; default effort moved from high to medium\n- Positioning: **Claude Fable 5.1-level results on most work** at less than half its price ($10\u002F$50)\n- Knowledge cutoff: June 2026\n\n## Availability and Compliance\n\n- Claude API, Amazon Bedrock, Google Cloud Vertex AI, Microsoft Azure\u002FFoundry, and Claude Pro\u002FMax\u002FTeam plans\n- Externally tested by METR and Frontier Design; ships with watermarking for EU AI Act compliance","BenchLM（Claude API Pricing 汇总）","https:\u002F\u002Fbenchlm.ai\u002Fanthropic\u002Fapi-pricing","2026-10-01T05:14:10.732Z"]