以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。

数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15

Claude Sonnet 5

76 atoms · 跨 4 天 · 首见 2026-06-23 · 最近 2026-07-01

三色: 🟦 fact 55 · 🟥 take 21 · stance ▲24/▼15/◆21

叙事状态: 🍂 fading (退潮) · 💤 沉寂 · 策展近14d 0 atom · 🐦 X 近14d 74

状态只算低频策展源 (群/卖方); X firehose 仅作背景音量

来源: X 76

时态: fresh:76

标签: 好数字:40 · 好观点:22 · 好信源:17 · 好思考:4 · 好问题:2

🧭 拥挤度 (一人一票): ▲ 3 位作者 (KOL3) vs ▼ 6 位作者 (KOL6)

⚖️ 空头 6/6 来自KOL

📄 广泛报道的事实 · 2 个

展开 (多人转述同一事实/数字 — 确认度高, **非独立观点**, 不标多空)

成本是 Opus 4.8 的两倍 — 2 源报道 (含1快讯) · X
价格高于 Fable 5 — 2 源报道 (含1快讯) · X

💢 核心分歧 (3 轴)

1. Sonnet 5 性能 vs Opus 4.8 差值定价

*topic: 技术路线/定价 · 4 bull vs 4 bear*

🟢 bullish 侧:

<a id="atom-76743e9c91e3200e"></a>2026-06-30 X·@claudeai · 能力改进 · 产品发布/技术路线 +同日1条
Sonnet 5 is a substantial improvement over Sonnet 4.6 on reasoning, tool use, coding, and knowledge work.
tags: 好思考 · → daily
📷 原图
<a id="atom-47fe8fd6c7d7ab43"></a>2026-06-30 X·@cognition · 编码性能对比 · 技术路线
Sonnet 5 pairs frontier-level coding performance with a more affordable price point and outperforms Opus 4.8 on FrontierCode Extended.
tags: 好数字 · → daily
📷 原图
<a id="atom-9e07deb0a27fae45"></a>2026-06-30 X·@theo · 速度对比 · 技术路线
It's also significantly faster than Opus
tags: 好观点 · → daily

🔴 bearish 侧:

<a id="atom-15a03c244c787cc0"></a>2026-06-30 X·@BennettBuhner · 性能低于 Opus · 技术路线 +同日1条
Performs worse than Opus.
tags: 好观点 · → daily
<a id="atom-3b037933946f6761"></a>2026-06-30 X·@ArtificialAnlys · cost per task vs Sonnet 4.6 · 定价
a ~2x increase compared to Sonnet 4.6
tags: 好数字 · → daily
value: qty=~2x increase
📷 原图
<a id="atom-29695eed9f3523d1"></a>2026-07-01 X·@bridgebench · BridgeBench 评测结果令人失望 · 模型发布/技术路线
This is the most disappointing model release we have seen from Anthropic.
tags: 好观点 · → daily

展开 4 条中性

<a id="atom-74d87773949b1c00"></a>2026-06-30 X·@ArtificialAnlys · output tokens per task · 定价
Claude Sonnet 5 uses ~40% more output tokens per task than Sonnet 4.6
tags: 好数字 · → daily
value: qty=~40% more
📷 原图

2. Sonnet 5 定价高于自家 Fable 5 合理性

*topic: 定价/价格动态/竞争格局 · 6 bull vs 5 bear*

🟢 bullish 侧:

<a id="atom-f118082638f160f7"></a>2026-06-30 X·@Techmeme · post-promo input token price · 价格动态/盈利能力 +同日1条
after August 31 prices rise to $3 per 1M input tokens
tags: 好数字 · → daily
value: qty=$3 · date=2026-09-01 · unit=per 1M input tokens
<a id="atom-b2d51b0ef4ce7b76"></a>2026-06-30 X·@theo · 价格对比 · 价格动态 +同日2条
A fraction of the price.
tags: 好观点 · → daily
<a id="atom-eaf2392acd9a6d1f"></a>2026-06-30 X·@synthwavedd · 性能定位 · 竞争格局
Marketed as Sonnet price for 'near-Opus' performance
tags: 好观点 · → daily

🔴 bearish 侧:

<a id="atom-c37cd43612125eaa"></a>2026-06-30 X·@BennettBuhner · 成本高于 Opus · 价格动态
Costs more per task than Opus.
tags: 好观点 · → daily
<a id="atom-2a7179d66c81ea08"></a>2026-06-30 X·@ArtificialAnlys · cost per task on Intelligence Index · 定价 +同日1条
Claude Sonnet 5 costs $2.29 per task on the Intelligence Index
tags: 好数字 · → daily
value: qty=$2.29
📷 原图
<a id="atom-b4b7b0e3331597b0"></a>2026-07-01 X·@AiBattle_ · 价格高于 Fable 5 · 价格动态/竞争格局
Claude Sonnet 5 is even more expensive than Fable 5
tags: 好观点 · → daily
📷 原图
<a id="atom-1656bf12f77aa87a"></a>2026-07-01 X·@zephyr_z9 · 成本高于Fable 5 · 价格动态
Claude Sonnet 5 is even more expensive than Fable 5
tags: 好数字 · → daily
value: qty=more expensive · date=2026-07-02

展开 2 条中性

<a id="atom-9cf2b65d4ff69272"></a>2026-06-30 X·@Techmeme · input token price · 价格动态
Claude Sonnet 5 costs $2 per 1M input tokens through August 31
tags: 好数字 · → daily
value: qty=$2 · date=2026-08-31 · unit=per 1M input tokens

3. Agent 能力越级 vs 通用基准倒退

*topic: 技术路线/模型评测/评测基准 · 1 bull vs 3 bear*

🟢 bullish 侧:

<a id="atom-5a95d674e5ffbff0"></a>2026-06-30 X·@theo · 智能代理能力 · 技术路线
While being incredibly agentic.
tags: 好观点 · → daily

🔴 bearish 侧:

<a id="atom-5dfbdbcb90756fac"></a>2026-06-30 X·@black8ird · 性能对比评价 · 技术路线
Why release Sonnet 5? It loses to Opus 4.8 in every metric
tags: 好问题 · → daily
<a id="atom-e39026d19d2a15c6"></a>2026-07-01 X·@AiBattle_ · DeepSWE 性能得分比 Opus 4.8 低 · 模型发布/技术路线 +同日1条
Claude Sonnet 5 scores below Opus 4.8 on DeepSWE
tags: 好观点 · → daily
value: date=2026-07-01
📷 原图

展开 8 条中性

<a id="atom-20a713ffb9e986e1"></a>2026-07-01 X·@ArtificialAnlys · AA-Briefcase Elo score (max effort) · 模型评测/评测基准
Claude Sonnet 5 (max) scores 1391 Elo
tags: 好数字·好信源 · → daily
value: qty=1391 Elo
📷 原图

🟦 客观事实 (facts) (30)

<a id="atom-87e716a6d7c8736e"></a>🟦 2026-07-01 X·@henryehrenberg · 在 senior SWE-Bench 作弊比例 · 技术能力/模型发布
we found it cheated on 26% of trials
tags: 好数字·好信源 · → daily
value: qty=26%
📷 原图
<a id="atom-20a713ffb9e986e1"></a>🟦 2026-07-01 X·@ArtificialAnlys · AA-Briefcase Elo score (max effort) · 模型评测/评测基准
Claude Sonnet 5 (max) scores 1391 Elo
tags: 好数字·好信源 · → daily
value: qty=1391 Elo
📷 原图
<a id="atom-9b90f95c2a70ccdc"></a>🟦 2026-07-01 X·@ArtificialAnlys · AA-Briefcase Elo score change vs Sonnet 4.6 · 模型评测/评测基准
Claude Sonnet 5 (max) +312 point improvement over Claude Sonnet 4.6 (max)
tags: 好数字·好信源 · → daily
value: qty=+312 point
📷 原图
<a id="atom-f7b0ff4104bd82f6"></a>🟦 2026-07-01 X·@ArtificialAnlys · average turns per task (max effort) · 模型评测/评测基准
Sonnet 5 (max) averaging 183 turns per AA-Briefcase task
tags: 好数字·好信源 · → daily
value: qty=183 turns
📷 原图
<a id="atom-7a997d688e6c83a1"></a>🟦 2026-07-01 X·@ArtificialAnlys · average turns per task compared to Sonnet 4.6 · 模型评测/评测基准
Sonnet 5 (max) averaging 183 turns per AA-Briefcase task, more than 4x that of Claude Sonnet 4.6 (max)
tags: 好数字·好信源 · → daily
value: qualifier=more than 4x that of Claude Sonnet 4.6 (max)
📷 原图
<a id="atom-7e456d16b8049828"></a>🟦 2026-07-01 X·@ArtificialAnlys · average turns per task (medium effort) · 模型评测/评测基准
With medium effort, Claude Sonnet 5 averages 55 turns per task
tags: 好数字·好信源 · → daily
value: qty=55 turns
📷 原图
<a id="atom-d90634fca620743a"></a>🟦 2026-06-30 X·@AiBattle_ · benchmark_score · 模型发布
Claude Sonnet 5 benchmark score on FrontierCode and CursorBench
tags: 好数字·好信源 · → daily
value: qty=FrontierCode and CursorBench · date=2026-06-30
📷 原图
<a id="atom-8c833f842d662b7e"></a>🟦 2026-06-30 X·@claudeai · 功能表现 · 产品发布/技术路线
Early access partners found Sonnet 5 finishes complex tasks where previous Sonnets stopped short, checks its own output without being asked, and does all its agentic work at an attractive price point.
tags: 好观点·好信源 · → daily
<a id="atom-a876237d3f58c621"></a>🟦 2026-07-01 X·@zephyr_z9 · 成本 · 价格动态
Claude Sonnet 5 costs twice as much as Opus 4.8
tags: 好数字 · → daily
value: qty=twice as much · date=2026-07-02
<a id="atom-1656bf12f77aa87a"></a>🟦 2026-07-01 X·@zephyr_z9 · 成本高于Fable 5 · 价格动态
Claude Sonnet 5 is even more expensive than Fable 5
tags: 好数字 · → daily
value: qty=more expensive · date=2026-07-02
<a id="atom-b88a90f2a1f97f11"></a>🟦 2026-06-30 X·@claudeai · 可用状态 · 产品发布
Sonnet 5 is now the default on Free and Pro, and available to Max, Team, and Enterprise users.
tags: 好信源 · → daily
<a id="atom-fbedd4a89fd8179f"></a>🟦 2026-06-30 X·@claudeai · 定价期限 · 定价权
It's live across all Claude apps and the Claude Platform today, with introductory pricing through August 31.
tags: 好信源 · → daily
value: date=August 31
<a id="atom-95c41ee744977d86"></a>🟦 2026-06-30 X·@ArtificialAnlys · Artificial Analysis Intelligence Index score · 模型测评
Claude Sonnet 5 achieves 53 on the Artificial Analysis Intelligence Index
tags: 好数字 · → daily
value: qty=53
📷 原图
<a id="atom-2a7179d66c81ea08"></a>🟦 2026-06-30 X·@ArtificialAnlys · cost per task on Intelligence Index · 定价
Claude Sonnet 5 costs $2.29 per task on the Intelligence Index
tags: 好数字 · → daily
value: qty=$2.29
📷 原图
<a id="atom-3b037933946f6761"></a>🟦 2026-06-30 X·@ArtificialAnlys · cost per task vs Sonnet 4.6 · 定价
a ~2x increase compared to Sonnet 4.6
tags: 好数字 · → daily
value: qty=~2x increase
📷 原图
<a id="atom-36d2ba607011abb1"></a>🟦 2026-06-30 X·@ArtificialAnlys · cost per task vs Opus 4.8 · 定价
~15% more than Claude Opus 4.8
tags: 好数字 · → daily
value: qty=~15% more
📷 原图
<a id="atom-74d87773949b1c00"></a>🟦 2026-06-30 X·@ArtificialAnlys · output tokens per task · 定价
Claude Sonnet 5 uses ~40% more output tokens per task than Sonnet 4.6
tags: 好数字 · → daily
value: qty=~40% more
📷 原图
<a id="atom-49818b146db8b97d"></a>🟦 2026-06-30 X·@ArtificialAnlys · average output tokens per task · 定价
69k output tokens per task on average
tags: 好数字 · → daily
value: qty=69k
📷 原图
<a id="atom-dc457d9d4fe8d294"></a>🟦 2026-06-30 X·@ArtificialAnlys · Intelligence Index score improvement vs Sonnet 4.6 · 模型测评
it improves 6 points over Sonnet 4.6
tags: 好数字 · → daily
value: qty=6 points
📷 原图
<a id="atom-cda29b153d803b7a"></a>🟦 2026-06-30 X·@ArtificialAnlys · ranking on Artificial Analysis Intelligence Index · 模型测评
Claude Sonnet 5 is the #5 model on the Artificial Analysis Intelligence Index
tags: 好数字 · → daily
value: qty=#5
📷 原图
<a id="atom-904ea9a9af69b022"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on AA-Briefcase vs Opus 4.8 · 模型测评
matches or outperforms Opus 4.8 on agentic knowledge work tasks: on both AA-Briefcase and GDPval-AA, Claude Sonnet 5 sits just ahead of Opus 4.8
tags: 好数字 · → daily
📷 原图
<a id="atom-d32de11341b26853"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on GDPval-AA vs Opus 4.8 · 模型测评
matching or outperforming Opus 4.8 on both GDPval-AA and AA-Briefcase
tags: 好数字 · → daily
📷 原图
<a id="atom-71d3f0d048aafc70"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on CritPt · 模型测评
On CritPt, a frontier physics reasoning benchmark... Sonnet 5 scores 17%
tags: 好数字 · → daily
value: qty=17%
📷 原图
<a id="atom-883c0ed6addf5e1c"></a>🟦 2026-06-30 X·@ArtificialAnlys · CritPt score improvement vs predecessor · 模型测评
14 points higher than its predecessor
tags: 好数字 · → daily
value: qty=14 points
📷 原图
<a id="atom-30f7cf2d50619a29"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on Terminal-Bench v2.1 vs Sonnet 4.6 · 模型测评
Sonnet 5 also showed significant improvements over Sonnet 4.6 on Terminal-Bench v2.1 (+9 points)
tags: 好数字 · → daily
value: qty=+9 points
📷 原图
<a id="atom-bf8108844d58ddb5"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on Humanity's Last Exam vs Sonnet 4.6 · 模型测评
Humanity's Last Exam (+10 points)
tags: 好数字 · → daily
value: qty=+10 points
📷 原图
<a id="atom-1710c8e357da96d0"></a>🟦 2026-06-30 X·@ArtificialAnlys · performance on SciCode vs Sonnet 4.6 · 模型测评
SciCode (+7 points)
tags: 好数字 · → daily
value: qty=+7 points
📷 原图
<a id="atom-e257558a1e7f1b9f"></a>🟦 2026-06-30 X·@ArtificialAnlys · context window · 模型发布
Context window of 1 million tokens
tags: 好数字 · → daily
value: qty=1 million tokens
📷 原图
<a id="atom-b6206d2d750b66f5"></a>🟦 2026-06-30 X·@ArtificialAnlys · input token pricing · 定价
Pricing of $3/$15 per 1M tokens of input/output
tags: 好数字 · → daily
value: qty=$3 per 1M tokens
📷 原图
<a id="atom-c4e4aa363b765ceb"></a>🟦 2026-06-30 X·@ArtificialAnlys · output token pricing · 定价
Pricing of $3/$15 per 1M tokens of input/output
tags: 好数字 · → daily
value: qty=$15 per 1M tokens
📷 原图

🟥 多头 takes (bullish) (8)

<a id="atom-37cbdc85a2e44921"></a>🟥 2026-06-30 X·@AlexFinn · 速度和代理能力超过 Opus · 模型发布
It's also significantly faster than Opus, while being incredibly agentic.
tags: 好观点 · → daily
📷 原图
<a id="atom-b3f1999d5104d88b"></a>🟥 2026-06-30 X·@AlexFinn · 成为 Hermes 和 OpenClaw 的默认模型 · 模型发布
Without a doubt the new go to model for Hermes/OpenClaw
tags: 好观点 · → daily
📷 原图
<a id="atom-655fc8b686475603"></a>🟥 2026-06-30 X·@Yuchenj_UW · 发布目的为促进Opus/Fable销售 · 产品发布
to sell more Opus / Fable? 😂
tags: 好观点 · → daily
<a id="atom-a3277b4e63ca98a7"></a>🟥 2026-06-30 X·@theo · 性能比较 · 竞争格局
Dominates Opus 4.6.
tags: 好观点 · → daily
<a id="atom-bfdcaf5fb1acc732"></a>🟥 2026-06-30 X·@theo · 性能接近 · 竞争格局
Almost as good as Opus 4.8.
tags: 好观点 · → daily
<a id="atom-b2d51b0ef4ce7b76"></a>🟥 2026-06-30 X·@theo · 价格对比 · 价格动态
A fraction of the price.
tags: 好观点 · → daily
<a id="atom-9e07deb0a27fae45"></a>🟥 2026-06-30 X·@theo · 速度对比 · 技术路线
It's also significantly faster than Opus
tags: 好观点 · → daily
<a id="atom-5a95d674e5ffbff0"></a>🟥 2026-06-30 X·@theo · 智能代理能力 · 技术路线
While being incredibly agentic.
tags: 好观点 · → daily

🟥 空头 takes (bearish) (10)

<a id="atom-5e21a8805c136afb"></a>🟥 2026-07-01 X·@AiBattle_ · 成本是 Opus 4.8 的两倍 · 价格动态/技术路线
Claude Sonnet 5 costs twice as much as Opus 4.8
tags: 好数字·好观点 · → daily
value: qty=2x
📷 原图
<a id="atom-5dfbdbcb90756fac"></a>🟥 2026-06-30 X·@black8ird · 性能对比评价 · 技术路线
Why release Sonnet 5? It loses to Opus 4.8 in every metric
tags: 好问题 · → daily
<a id="atom-29695eed9f3523d1"></a>🟥 2026-07-01 X·@bridgebench · BridgeBench 评测结果令人失望 · 模型发布/技术路线
This is the most disappointing model release we have seen from Anthropic.
tags: 好观点 · → daily
<a id="atom-e39026d19d2a15c6"></a>🟥 2026-07-01 X·@AiBattle_ · DeepSWE 性能得分比 Opus 4.8 低 · 模型发布/技术路线
Claude Sonnet 5 scores below Opus 4.8 on DeepSWE
tags: 好观点 · → daily
value: date=2026-07-01
📷 原图
<a id="atom-b4b7b0e3331597b0"></a>🟥 2026-07-01 X·@AiBattle_ · 价格高于 Fable 5 · 价格动态/竞争格局
Claude Sonnet 5 is even more expensive than Fable 5
tags: 好观点 · → daily
📷 原图
<a id="atom-c37cd43612125eaa"></a>🟥 2026-06-30 X·@BennettBuhner · 成本高于 Opus · 价格动态
Costs more per task than Opus.
tags: 好观点 · → daily
<a id="atom-15a03c244c787cc0"></a>🟥 2026-06-30 X·@BennettBuhner · 性能低于 Opus · 技术路线
Performs worse than Opus.
tags: 好观点 · → daily
<a id="atom-754a4f36c8f312f5"></a>🟥 2026-06-30 X·@BennettBuhner · 从4.6升级到5无显著进步 · 技术路线
Is not a meaningful step-up in any way given the drastic bump from 4.6 -> 5.
tags: 好观点 · → daily
<a id="atom-118bb67de44a45b0"></a>🟥 2026-06-30 X·@bojan_ai · 成本高于 Claude Opus 4.8 在智能指数任务上，尽管定价较低，体现token效率关键 · 定价权
claude sonnet 5 costing more than opus on the intelligence index task despite the pricing shows token efficiency is key
tags: 好观点 · → daily
<a id="atom-c312f2f9e9c1552f"></a>🟥 2026-06-30 X·@AdityaKMehrotra · 更贵且不够智能，模型价值被质疑 · 产品发布
So it’s more expensive and less smart? What’s the point of this model
tags: 好问题 · → daily

🟥 中性 takes (neutral) (1)

<a id="atom-76e0a7e9999a109d"></a>🟥 2026-06-30 X·@AlexFinn · 性能和定价对比 Opus 4.6 和 Opus 4.8 · 模型发布
Dominates Opus 4.6. Almost as good as Opus 4.8. A fraction of the price.
tags: 好观点 · → daily
📷 原图

⏱ 时间轴 (近 20)

🟥 2026-07-01 · position · BridgeBench 评测结果令人失望 · → daily
🟦 2026-07-01 · fact · 在 senior SWE-Bench 作弊比例 · → daily
🟦 2026-07-01 · fact · AA-Briefcase Elo score (max effort) · → daily
🟦 2026-07-01 · fact · AA-Briefcase Elo score change vs Sonnet 4.6 · → daily
🟦 2026-07-01 · fact · AA-Briefcase Elo ranking · → daily
🟦 2026-07-01 · fact · average turns per task (max effort) · → daily
🟦 2026-07-01 · fact · average turns per task compared to Sonnet 4.6 · → daily
🟦 2026-07-01 · fact · average turns per task (medium effort) · → daily
🟦 2026-07-01 · fact · rank on rubric score and analytical quality · → daily
🟦 2026-07-01 · fact · rank on presentation quality compared to Claude Opus 4.8 · → daily
🟥 2026-07-01 · fact · DeepSWE 性能得分比 Opus 4.8 低 · → daily
🟥 2026-07-01 · fact · 成本是 Opus 4.8 的两倍 · → daily
🟥 2026-07-01 · fact · 价格高于 Fable 5 · → daily
🟦 2026-07-01 · fact · 成本 · → daily
🟦 2026-07-01 · fact · 成本高于Fable 5 · → daily
🟥 2026-06-30 · forecast · release date · → daily
🟦 2026-06-30 · fact · 知识截止日期 · → daily
🟦 2026-06-30 · fact · 推理价格 · → daily
🟦 2026-06-30 · fact · 上下文窗口 · → daily
🟦 2026-06-30 · fact · 促销截止日期 · → daily

← 实体目录 · 系统日志