大模型
www.aibase.com ·
Anthropic 抢在 OpenAI 开发者大会前发布 Claude Sonnet 5.5：半价逼近 Opus，并通关《宝可梦红》
AIbase 报道称，Anthropic 在 OpenAI 年度开发者大会前发布 Claude Sonnet 5.5，定位日常通用助手，速度较 Sonnet 5 提升 30%，单任务成本最多降低 30%，定价为 Opus 5.5 的一半，并在 Terminal-Bench 4.0 等测试中接近甚至超过自家高端模型。文章还称该模型首次仅凭截图完成《宝可梦红》，体现视觉理解与长期规划能力，并支持五档 Effort Level 调节，最高档可能因输出 token 激增导致成本反超 Opus 5.5 且性能下降。此外提到其加入反蒸馏机制和与 Opus 同级的网络安全防护。上述信息来自聚合报道，缺少官方来源和可验证数据。
AI推荐理由 内容涉及 Anthropic 在 OpenAI 开发者大会前发布 Claude Sonnet 5.5，给出价格、速度、成本、Terminal-Bench 4.0 表现和通关《宝可梦红》等具体信息，对模型选型和 Agent 能力评估有参考价值。但来源为 AIbase 聚合站，正文未提供官方链接、基准数据表或可验证证据，'性能几乎匹配 Opus'、'通关宝可梦'等关键结论无法独立核实，且带有明显营销式表述，需谨慎对待。
62
正文
Anthropic released Claude Sonnet 5.5 before OpenAI's Developer Conference, half the price of Opus and even passed Pokémon
AIbase基地
Published inAI News · 3 min read · Sep 29, 2026
5
Anthropic released Claude Sonnet 5.5 just in time for OpenAI's annual developer conference. This new model, positioned as a versatile daily assistant, breaks the so-called impossible triangle of speed, low cost, and high quality—speed increased by 30% compared to the previous Sonnet 5, single-task cost reduced by up to 30%, priced at half of Opus 5.5, with performance almost matching, and even surpassing its own premium models in several tests like Terminal-Bench 4.0.
The most convincing proof of its intelligence is its performance in games. Sonnet 5.5 became the first Sonnet model to complete "Pokémon Red" by only watching screenshots—without using memory modification or built-in guides, it understood the world, made decisions, and advanced the story frame by frame, integrating visual understanding and long-term planning into one brain.
Its capabilities go far beyond coding. It performs exceptionally well in scenarios such as office work, design, and 3D simulation, and supports five levels of Effort Level adjustment: low and medium levels are ideal for routine tasks, offering excellent value for money; however, when the highest level is fully activated, the output token volume increases dramatically, potentially making the cost exceed that of the larger Opus 5.5, and even leading to a performance drop—meaning higher levels aren't always better; you need to adjust based on the workload.
The moat has been further strengthened. Sonnet 5.5 includes anti-distillation mechanisms and features cybersecurity protection at the same level as Opus, preventing others from learning from its model while ensuring security standards match those of flagship products. When Anthropic released Sonnet 5.5 at half the price, with performance on par and even a real-world demonstration of completing Pokémon, it clearly aimed not only at developers' wallets but also at pricing power in the entire mid-range market.
ClaudeSonnet5.5AI New WordsBrand Product TermsSonnet5.5
This article is from AIbase Daily
Scan to view
Welcome to the [AI Daily] column! This is your daily guide to exploring the world of artificial intelligence. Every day, we present you with hot topics in the AI field, focusing on developers, helping you understand technical trends, and learning about innovative AI product applications.
—— Created by the AIbase Daily Team
© Copyright AIbase Base 2024, Click to View Source -