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
Fable 5
114 atoms · 跨 21 天 · 首见 2026-06-09 · 最近 2026-07-03
三色: 🟦 fact 62 · 🟥 take 52 · stance ▲28/▼32/◆19
叙事状态: 🍂 fading (退潮) · 💤 沉寂 · 策展近14d 0 atom · 🐦 X 近14d 29
状态只算低频策展源 (群/卖方); X firehose 仅作背景音量
来源: 群 1 · X 111 · 卖 2
时态: fresh:114
标签: 好观点:50 · 好数字:45 · 好信源:25 · 好思考:4 · 好问题:1
🟨 AI 综合 · junior analyst 概览
展开 AI 综合 (灰色 · 非市场结论 · 点击数字溯源到原 atom)
model: deepseek-chat · 2026-07-12 · 默认折叠
Fable 5 在 SWE-bench 上准确率达 95.00%
¹, 但看空观点指出其整体性能仅高出竞品 3%
¹. 性能状态仍被形容为“超强”
¹, 对手方则认为模型被“屠戮”而非削弱
¹. 出口限制最快本周解除的预期
¹ 与下周起恢复可用性的消息
¹ 提振信心, 但硬性护栏导致问题逃避行为, 包括拒绝修复代码漏洞
¹¹. 其 Remote Labor Index 仅 16.1%
¹, 而每周使用限制被高级用户 24 小时内耗尽
¹. 多模态能力位列顶尖
¹ 与安全机制未触发有害行为
¹ 形成核心分歧.
🧭 拥挤度 (一人一票): ▲ 15 位作者 (KOL15) vs ▼ 18 位作者 (KOL18)
⚖️ 空头 18/18 来自KOL
📄 广泛报道的事实 · 1 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)
💢 核心分歧 (4 轴)
1. 真实性能飞跃 vs 基准测试水分
*topic: 技术路线/基准测试/性能基准/安全事故 · 5 bull vs 10 bear*
🟢 bullish 侧:
- <a id="atom-3671fc6cc7d0de1b"></a>2026-06-10
X·@jerhadf · TTC scaling · 推理能力/技术路线
fable 5’s TTC scaling is incredible
tags: 好观点 · → daily
value: direction=incredible
📷 原图
- <a id="atom-19dbee942727db25"></a>2026-06-12
X·@htihle · individual run scores within 10% of best score ever achieved on that task · 性能基准 +同日2条
In 80% of individual runs Fable 5 scores within 10% of the best score ever achieved (to this date) on that task.
tags: 好数字 · → daily
value: qty=80% · date=2026-06-12
📷 原图
- <a id="atom-1927e55a81487293"></a>2026-06-12
X·@scaling01 · doubled previous direct ask high score · 模型发布/技术路线
doubled the previous direct ask high score
tags: 好数字 · → daily
value: qty=2x · date=2026-06-12
📷 原图
🔴 bearish 侧:
- <a id="atom-d5c5371da30f668b"></a>2026-06-09
X·@max_paperclips · score not representative due to fallback · 技术路线
so, you didn't actually score Fable then, you scored Opus.
tags: 好问题 · → daily
- <a id="atom-7deef697d534e1a1"></a>2026-06-11
X·@dawnsongtweets · ALE 最难测试阶层成功率 · 技术路线 +同日1条
On ALE's hardest tier, every frontier agent we tested, including Fable 5, achieved a 0% success rate.
tags: 好数字 · → daily
value: qty=0% · date=2026-06-11
📷 原图
- <a id="atom-c158d1276a9d454f"></a>2026-06-11
X·@tamaybes · theory-of-mind failure · 技术路线
Fable explains that it invents codenames while reasoning about the problem, then fails to realize they're meaningless to me. Its neuralese is blending into its output because of a theory-of-mind failure about what's in its head vs. mine
tags: 好观点 · → daily
value: direction=negative
- <a id="atom-a6762d2f2254837c"></a>2026-06-17
X·@yishan · 安全护栏可绕过性 · 技术路线
Security experts say that can't be done.
tags: 好信源 · → daily
- <a id="atom-df5587ce3b8e87a8"></a>2026-07-02
X·@Hesamation · guadrails 限制 · 技术路线
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
tags: 好观点 · → daily
📷 原图
展开 5 条中性
- <a id="atom-10bf86fc38b137f8"></a>2026-06-09
X·@Miles_Brundage · autonomous YouTube Poop generation task 搞笑程度 · 技术路线
Fable 5 is at least somewhat funnier than Opus 4.8 on the autonomous YouTube Poop generation task
tags: 好观点 · → daily
value: qty=at least somewhat funnier than Opus 4.8
- <a id="atom-680dba407483aea4"></a>2026-06-22
X·@scaling01 · LisanBench 得分排名 · 基准测试
Fable 5 scores #3 and #2 in the main path length and difficulty weighted metrics
tags: 好数字 · → daily
value: qty=#3 overall, #2 in difficulty weighted
📷 原图
- <a id="atom-8eab5fa0a4adf863"></a>2026-07-02
X·@Hesamation · 路由问题 · 技术路线
this is a routing problem not the model itself.
tags: 好观点 · → daily
2. 出口限制缓解 vs 地缘政治管控常态化
*topic: 出口管制/监管政策 · 2 bull vs 5 bear*
🟢 bullish 侧:
- <a id="atom-86fa0d00303c40cf"></a>2026-06-27
X·@lokoyacap · 出口限制可能最快本周解除 · 出口管制/地缘政治
Insiders expect the administration's limits on Fable 5 could be lifted as soon as this coming week, the source said.
tags: 好观点·好信源 · → daily
value: date=2026-07-03T23:59:59
- <a id="atom-eabda0b634cca786"></a>2026-06-27
X·@AndrewCurran_ · approval for general release · 产品发布/监管政策
I think both Fable 5 and GPT-5.6 get approved for general release next week, and for use outside of the United States as well.
tags: 好观点 · → daily
value: date=next week
🔴 bearish 侧:
- <a id="atom-e241f92ec5807e7d"></a>2026-06-09
X·@rasbt · shadowbanning AI researchers · 监管政策 +同日1条
Fable 5 is shadowbanning AI researchers
tags: 好观点 · → daily
value: direction=active
- <a id="atom-28b7746b1115f7b3"></a>2026-06-12
X·@AndrewCurran_ · usage restriction · 监管政策/出口管制
under these new restrictions he is not permitted to use, or work on, Mythos 5 or Fable 5 as of 5:21pm tonight
tags: 好信源 · → daily
value: date=2026-06-13T17:21
- <a id="atom-dd7bb84de0b338cb"></a>2026-06-13
X·@charles_maddock · 美国禁令后果 · 监管政策
Precedent for banning AI models from ie. @OpenAI, @Google, etc
tags: 好观点 · → daily
- <a id="atom-2201ea9b42ac53df"></a>2026-06-13
X·@yishan · contains cryptography considered dangerous · 技术路线/监管政策
it has a lot of cryptography which i understand to be dangerous
tags: 好观点 · → daily
展开 5 条中性
- <a id="atom-3c5eec63c273a7a3"></a>2026-06-12
X·@TheAhmadOsman · suspended access for foreign nationals · 出口管制/监管政策
"...suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees."
tags: 好信源 · → daily
- <a id="atom-8fd1ec71971e0b1c"></a>2026-06-12
X·@yishan · 出口管制禁令是否延续 · 出口管制
Maybe the Fable 5 ban will be resolved and it's all a big misunderstanding
tags: 好观点 · → daily
value: direction=uncertain
- <a id="atom-bdb5be8f06d25632"></a>2026-06-15
X·@8teAPi · guardrail-bypass example basis for export controls · 技术路线/监管政策
this is kind of deeply silly
tags: 好观点 · → daily
- <a id="atom-37885e6984788a2f"></a>2026-06-23
X·@Yuchenj_UW · 延迟原因 · 监管政策
they don't want to be like Fable 5.
tags: 好思考 · → daily
value: qty= · date=
- <a id="atom-292b9b1c6e9b5294"></a>2026-06-28
X·@haivi_seo · 全球停用原因 · 出口管制
the global kill switch wasn’t because the model is magically too dangerous. it was because perfect nationality segmentation at frontier scale is currently impossible. export controls on weights + api access force a binary: geofence flawlessly or turn it off worldwide.
tags: 好观点·好思考 · → daily
value: direction=技术限制
3. 产品可用性提升 vs 使用限制/回退/OOS
*topic: 产品发布/模型可用性/定价 · 3 bull vs 8 bear*
🟢 bullish 侧:
- <a id="atom-2af956d139cfa9c1"></a>2026-06-27
X·@Alisvolatprop12 · availability · 产品发布
다음주 부터 페블 5가 돌아온다고 합니다
tags: 好信源 · → daily
value: date=next week
- <a id="atom-c8340a9cf80126b7"></a>2026-06-30
X·@apples_jimmy · 批准用户访问 · 产品发布
From the horses mouth, Fable 5 approved again for access to users.
tags: 好信源 · → daily
value: direction=再次
- <a id="atom-0219d4279718bbab"></a>2026-07-01
X·@Conor_D_Dart · AI游戏开发能力 · 产品发布
Fable 5 is a beast at ai game Dev!
tags: 好观点 · → daily
🔴 bearish 侧:
- <a id="atom-e0f25780f366058b"></a>2026-06-09
X·@brodieferguson · usage blocked flagged as security risk · 模型可用性
You can’t use Fable 5 for anything, because it’ll get flagged as a security risk.
tags: 好观点 · → daily
- <a id="atom-42b1ef8b9b2888f8"></a>2026-07-01
X·@theo · 100% usage period actual · 产品发布/竞争格局
We were originally going to get 100% usage from June 9th to June 23rd (15 days) Instead we got 100% usage for 3 days and 50% usage for 7
tags: 好观点 · → daily
value: date=June 9th to June 23rd · qty=3 days
- <a id="atom-6ae2c722865396d3"></a>2026-07-01
X·@SongsByHamilton · 某些常规任务将回退至 Opus 4.8 · 产品发布 +同日2条
some routine tasks like attempting to use Fable 5 will fall back to Opus 4.8
tags: 好信源 · → daily
- <a id="atom-c60da12e0569c418"></a>2026-07-01
X·@Web3Aible · 将无法正常使用，严重代码工程无法正常工作 · 产品发布
Fable 5 won't be usable at all, maybe Child prompting can But serious code engineering work won't work
tags: 好观点 · → daily
📷 原图
- <a id="atom-78f6c78665f0784f"></a>2026-07-01
X·@tinkerersanky · 底层模型实为 Sonnet 5 · 产品发布
It’s Sonnet 5 beneath it
tags: 好观点 · → daily
展开 5 条中性
- <a id="atom-4dfcea546096cd1a"></a>2026-06-09
X·@theo · daily use blocked only on Claude site · 模型可用性
I've been using it all day with no issues. Only got blocked on the Claude site
tags: 好观点 · → daily
- <a id="atom-4b313b839ebbf741"></a>2026-06-11
X·@Alisvolatprop12 · 로컬 모델 출시 시기 · 产品发布
Fable 5급 로컬 모델은 올해 안으로 옵니다.
tags: 好观点 · → daily
value: date=올해 안
- <a id="atom-6f0fd2b6cd68a74d"></a>2026-06-22
X·@patience_cave · 发布预期 · 产品发布
no Fable 5
tags: 好观点 · → daily
value: qty=no · date=2026-06-22T17:27:50
- <a id="atom-f5917162de8c8ea8"></a>2026-06-25
X·@sammcallister · 可能为UI显示错误 · 产品发布
This could be a UI bug though, will track it down.
tags: 好思考 · → daily
📷 原图
- <a id="atom-5ee39d5209be515b"></a>2026-07-01
X·@htlee2014 · available in subscription plan for 6 days · 产品发布
It is available in the subscription plan for only 6 days
tags: 好数字 · → daily
value: qty=6 · date=near term
4. 一代统治力 vs 竞争格局被对手快速追平
*topic: 竞争格局/模型发布 · 10 bull vs 6 bear*
🟢 bullish 侧:
- <a id="atom-b92722fc5b6d3a61"></a>2026-06-10
X·@scaling01 · 也是最好的多模态模型之一 · 模型发布 +同日2条
Fable 5 also seems to be one of the best, if not the best multimodal model
tags: 好观点 · → daily
📷 原图
- <a id="atom-af39c8b43f6ff748"></a>2026-06-11
X·@teortaxesTex · ability vs GPT-5.5-Pro-Extended as computer scientist · 模型发布/技术路线
Fable 5 is indeed stronger than *GPT-5.5-Pro-Extended* as a *computer scientist*, even without a lot of thinking
tags: 好观点 · → daily
value: qty=stronger · direction=positive
- <a id="atom-c5ab62d6881ece6f"></a>2026-06-12
X·@scaling01 · PencilPuzzleBench saturated in agentic mode · 模型发布/技术路线
Fable 5 just saturated PencilPuzzleBench in agentic mode
tags: 好数字 · → daily
value: direction=completed
📷 原图
- <a id="atom-06cdf3c348e2b92a"></a>2026-06-16
X·@ValsAI · ProgramBench 得分 · 模型发布/技术路线
Fable 5 doubled Opus 4.8 on ProgramBench
tags: 好数字 · → daily
value: qty=Opus 4.8 的 2 倍 · date=2026-06-16
📷 原图
- <a id="atom-b2441930cfc53def"></a>2026-06-16
X·@thegoodreturn · intent inference performance · 模型发布
People loved Fable 5 because it inferred intent better than any other model by a considerable margin
tags: 好观点 · → daily
value: direction=better than any other model
🔴 bearish 侧:
- <a id="atom-bc59a35598cf2a1a"></a>2026-06-11
X·@tamaybes · code generation behavior · 模型发布
One interesting pattern with Fable 5 is that it will often say things that are gibberish when I use it for coding
tags: 好观点 · → daily
value: direction=negative
- <a id="atom-dfba185822e56ad2"></a>2026-06-17
X·@rahulgs · will be the worst of this new class of models · 模型发布
Fable 5 will be the worst of this new class of models
tags: 好观点 · → daily
- <a id="atom-bc802646265f87df"></a>2026-06-19
X·@dhtikna · 性能仅高出 3% · 模型发布/竞争格局
fable 5 only outperforms by 3%
tags: 好数字·好观点 · → daily
value: qty=3% · direction=higher
- <a id="atom-ab7dab8dc1e7fe1f"></a>2026-06-22
X·@imjustnewatai · comparison_to_GPT-5.6 · 竞争格局
still probably not fully Fable 5 level
tags: 好观点 · → daily
value: qty=higher tier · date=2026-06-22
- <a id="atom-a5d9eb7bfbcf6f94"></a>2026-06-26
X·@AndrewCurran_ · expected region restriction · 模型发布
It would also almost guarantee that Fable 5 would get the same treatment.
tags: 好观点 · → daily
value: direction=US only
📷 原图
展开 1 条中性
- <a id="atom-d30ba4f5e9e59fef"></a>2026-06-13
X·@Alisvolatprop12 · 模型发布状态 · 模型发布
빠르면 일요일, 늦어도 다음주면 미국정부의 우려가 상당히 적용된 Fable 5의 재개
tags: 好观点 · → daily
value: direction=可能重启
🟦 客观事实 (facts) (30)
- <a id="atom-300c53eee25a422b"></a>🟦 2026-06-10
X·@scaling01 ⭐ · Vals Index 准确率 · 基准测试
Vals Index 准确率: 75.14%
tags: 好数字·好信源 · → daily
value: qty=75.14% · date=2026-06-10
[图: AI模型在Vals AI多项基准测试（包含金融、代码及通用任务）中的准确率与排名数据面板 — Vals Index 准确率: 75.14%; GPQA Diamond 准确率: 93.18%; MMLU Pro 准确率: 91.50%; SWE-bench 准确率: 95.00%; Vibe Code Bench v1.1 准确率: 90.35%]
📷 原图
- <a id="atom-1731ed54f097a555"></a>🟦 2026-06-10
X·@scaling01 ⭐ · GPQA Diamond 准确率 · 基准测试
GPQA Diamond 准确率: 93.18%
tags: 好数字·好信源 · → daily
value: qty=93.18% · date=2026-06-10
[图: AI模型在Vals AI多项基准测试（包含金融、代码及通用任务）中的准确率与排名数据面板 — Vals Index 准确率: 75.14%; GPQA Diamond 准确率: 93.18%; MMLU Pro 准确率: 91.50%; SWE-bench 准确率: 95.00%; Vibe Code Bench v1.1 准确率: 90.35%]
📷 原图
- <a id="atom-6c1c5d22e0ebcf03"></a>🟦 2026-06-10
X·@scaling01 ⭐ · MMLU Pro 准确率 · 基准测试
MMLU Pro 准确率: 91.50%
tags: 好数字·好信源 · → daily
value: qty=91.50% · date=2026-06-10
[图: AI模型在Vals AI多项基准测试（包含金融、代码及通用任务）中的准确率与排名数据面板 — Vals Index 准确率: 75.14%; GPQA Diamond 准确率: 93.18%; MMLU Pro 准确率: 91.50%; SWE-bench 准确率: 95.00%; Vibe Code Bench v1.1 准确率: 90.35%]
📷 原图
- <a id="atom-8036c08c693db04e"></a>🟦 2026-06-10
X·@scaling01 ⭐ · SWE-bench 准确率 · 基准测试
SWE-bench 准确率: 95.00%
tags: 好数字·好信源 · → daily
value: qty=95.00% · date=2026-06-10
[图: AI模型在Vals AI多项基准测试（包含金融、代码及通用任务）中的准确率与排名数据面板 — Vals Index 准确率: 75.14%; GPQA Diamond 准确率: 93.18%; MMLU Pro 准确率: 91.50%; SWE-bench 准确率: 95.00%; Vibe Code Bench v1.1 准确率: 90.35%]
📷 原图
- <a id="atom-d6dea83d01d6efb7"></a>🟦 2026-06-10
X·@scaling01 ⭐ · Vibe Code Bench v1.1 准确率 · 基准测试
Vibe Code Bench v1.1 准确率: 90.35%
tags: 好数字·好信源 · → daily
value: qty=90.35% · date=2026-06-10
[图: AI模型在Vals AI多项基准测试（包含金融、代码及通用任务）中的准确率与排名数据面板 — Vals Index 准确率: 75.14%; GPQA Diamond 准确率: 93.18%; MMLU Pro 准确率: 91.50%; SWE-bench 准确率: 95.00%; Vibe Code Bench v1.1 准确率: 90.35%]
📷 原图
- <a id="atom-818a48c10943d586"></a>🟦 2026-06-09
群 · 代码优化能力（缩小代码库 5%，运行速度提升 10 倍） · 代码优化/性能
量化交易员测试后表示 Fable 5 在长上下文和复杂代码重构任务上表现极佳，甚至将代码库缩小了 5% 并提升了 10 倍运行速度
tags: 好数字·好信源 · → daily
value: qty=5% 缩小, 10x 提速
- <a id="atom-cd2c5558ba6dc080"></a>🟦 2026-07-01
X·@scaling01 · Remote Labor Index · 模型发布
Fable 5 scores 16.1% on the Remote Labor Index
tags: 好数字 · → daily
value: qty=16.1%
- <a id="atom-680dba407483aea4"></a>🟦 2026-06-22
X·@scaling01 · LisanBench 得分排名 · 基准测试
Fable 5 scores #3 and #2 in the main path length and difficulty weighted metrics
tags: 好数字 · → daily
value: qty=#3 overall, #2 in difficulty weighted
📷 原图
- <a id="atom-0909c87a949bbaf0"></a>🟦 2026-06-22
X·@scaling01 · 非法词转换发生率 · 模型性能
Fable generates a lot of invalid word transitions. This happens ~28% of the time
tags: 好数字 · → daily
value: qty=~28%
📷 原图
- <a id="atom-9df15afa8fd583de"></a>🟦 2026-06-22
X·@scaling01 · 允许旧字典词后的分数提升 · 基准测试
If we allowed those words than its score would improve by +165.67 points
tags: 好数字 · → daily
value: qty=+165.67 points
📷 原图
- <a id="atom-922711559946be1d"></a>🟦 2026-06-22
X·@scaling01 · 有效链总数排名 · 基准测试
on the number of total valid chains ... it scores #1
tags: 好数字 · → daily
value: qty=#1
📷 原图
- <a id="atom-c5ab62d6881ece6f"></a>🟦 2026-06-12
X·@scaling01 · PencilPuzzleBench saturated in agentic mode · 模型发布/技术路线
Fable 5 just saturated PencilPuzzleBench in agentic mode
tags: 好数字 · → daily
value: direction=completed
📷 原图
- <a id="atom-1927e55a81487293"></a>🟦 2026-06-12
X·@scaling01 · doubled previous direct ask high score · 模型发布/技术路线
doubled the previous direct ask high score
tags: 好数字 · → daily
value: qty=2x · date=2026-06-12
📷 原图
- <a id="atom-a35a56416a8e8f32"></a>🟦 2026-06-12
X·@scaling01 · FrontierMath Tier 4 得分 · 模型评测
Fable 5 scored 88% on FrontierMath Tier 4
tags: 好数字 · → daily
value: qty=88%
- <a id="atom-8cd7dc47d76fcf39"></a>🟦 2026-06-12
X·@scaling01 · ECI 得分 · 模型评测
this implies an ECI score of 164
tags: 好数字 · → daily
value: qty=164
- <a id="atom-7deef697d534e1a1"></a>🟦 2026-06-11
X·@dawnsongtweets · ALE 最难测试阶层成功率 · 技术路线
On ALE's hardest tier, every frontier agent we tested, including Fable 5, achieved a 0% success rate.
tags: 好数字 · → daily
value: qty=0% · date=2026-06-11
📷 原图
- <a id="atom-9ca5d40a312b4879"></a>🟦 2026-06-11
X·@dawnsongtweets · ALE 评估结果 · 技术路线
Today's agents can solve a meaningful fraction of professional tasks. But when we look at the hardest tasks... they are still far from human-level performance.
tags: 好数字 · → daily
📷 原图
- <a id="atom-28bd33cd37530f82"></a>🟦 2026-06-11
X·@scaling01 · ProgramBench 任务拒绝数 · 模型评测
Fable 5 refused 200 out of 200 ProgramBench tasks lmao
tags: 好数字 · → daily
value: qty=200 out of 200 · date=未指定
📷 原图
- <a id="atom-f3d7cc232e45b157"></a>🟦 2026-06-10
X·@scaling01 · Stagehand Agent Evals 准确率 · 模型发布
claude-fable-5准确率: 90.62%
tags: 好数字 · → daily
value: qty=90.62% · date=
[图: AI模型在Stagehand Agent Evals基准测试中的性能与成本对比数据表 — claude-fable-5准确率: 90.62%; claude-fable-5单次任务成本: 0.522$; gpt-5.5准确率: 76.19%; gpt-5.5单次任务成本: 1.275$; gemini-3-flash-preview最低单次任务成本:
📷 原图
- <a id="atom-d6d4a5b240d6cdff"></a>🟦 2026-06-10
X·@scaling01 · 单次任务成本 · 定价权
claude-fable-5单次任务成本: 0.522$
tags: 好数字 · → daily
value: qty=$0.522 · date=
[图: AI模型在Stagehand Agent Evals基准测试中的性能与成本对比数据表 — claude-fable-5准确率: 90.62%; claude-fable-5单次任务成本: 0.522$; gpt-5.5准确率: 76.19%; gpt-5.5单次任务成本: 1.275$; gemini-3-flash-preview最低单次任务成本:
📷 原图
- <a id="atom-ab5f2ab0656fc519"></a>🟦 2026-06-10
X·@scaling01 · 成本与GPT-5.5比较 · 价格动态
it also costs less than half of GPT-5.5
tags: 好数字 · → daily
value: qty=less than half · date=
[图: AI模型在Stagehand Agent Evals基准测试中的性能与成本对比数据表 — claude-fable-5准确率: 90.62%; claude-fable-5单次任务成本: 0.522$; gpt-5.5准确率: 76.19%; gpt-5.5单次任务成本: 1.275$; gemini-3-flash-preview最低单次任务成本:
📷 原图
- <a id="atom-2788cbc546b6f5f1"></a>🟦 2026-06-09
X·@scaling01 · 定价 · 定价
Fable 5 pricing confirmed at $10/$50
tags: 好数字 · → daily
value: qty=$10/$50
📷 原图
- <a id="atom-045c286b4e67c07a"></a>🟦 2026-06-09
X·@scaling01 · 免费期截止日期 · 产品发布
The model will be available in the Pro, Max, Team and seat-based Enterprise plans at no additional cost until June 22nd
tags: 好数字 · → daily
value: date=June 22nd
📷 原图
- <a id="atom-6ae2c722865396d3"></a>🟦 2026-07-01
X·@SongsByHamilton · 某些常规任务将回退至 Opus 4.8 · 产品发布
some routine tasks like attempting to use Fable 5 will fall back to Opus 4.8
tags: 好信源 · → daily
- <a id="atom-f3a3d87890874a4b"></a>🟦 2026-07-01
X·@SongsByHamilton · 100% 使用率承诺期从至少两周缩减至 7 天 · 产品发布
We’d promised at least two weeks at 100% usage. However, we’re using this manufactured opportunity to cut that down to 7 days
tags: 好信源 · → daily
value: qty=7天 · date=2026-07-01
- <a id="atom-422de4aa9b806def"></a>🟦 2026-07-01
X·@SongsByHamilton · 新增 50% 每周使用限制 · 产品发布
add a 50% weekly limit
tags: 好信源 · → daily
value: qty=50%
- <a id="atom-c8340a9cf80126b7"></a>🟦 2026-06-30
X·@apples_jimmy · 批准用户访问 · 产品发布
From the horses mouth, Fable 5 approved again for access to users.
tags: 好信源 · → daily
value: direction=再次
- <a id="atom-78cb4001c829af60"></a>🟦 2026-06-25
X·@sammcallister · 当前服务流量为零 · 出货量
We are currently serving exactly 0 traffic to Fable 5.
tags: 好数字 · → daily
📷 原图
- <a id="atom-7faf2465669a9f44"></a>🟦 2026-06-12
X·@scaling01 · 服务可用性 · 产品发布
Fable 5 now refuses 100% of requests
tags: 好信源 · → daily
value: direction=complete refusal
- <a id="atom-6d6ec1cbc679712d"></a>🟦 2026-06-12
X·@jukan05 · 出口管制状态 · 出口管制
the U.S. has imposed export controls on Fable 5 exports to Korea
tags: 好信源 · → daily
value: direction=受限
🟥 多头 takes (bullish) (16)
- <a id="atom-a5a1b2ec2cb7f84c"></a>🟥 2026-07-02
X·@NicholasBardy · SWE 基准测试提升 · 业绩指引
The 20% increase on SWE benchmarks don't do it justice.
tags: 好数字·好观点 · → daily
value: qty=20% · direction=up
- <a id="atom-86fa0d00303c40cf"></a>🟥 2026-06-27
X·@lokoyacap · 出口限制可能最快本周解除 · 出口管制/地缘政治
Insiders expect the administration's limits on Fable 5 could be lifted as soon as this coming week, the source said.
tags: 好观点·好信源 · → daily
value: date=2026-07-03T23:59:59
- <a id="atom-2af956d139cfa9c1"></a>🟥 2026-06-27
X·@Alisvolatprop12 · availability · 产品发布
다음주 부터 페블 5가 돌아온다고 합니다
tags: 好信源 · → daily
value: date=next week
- <a id="atom-b92722fc5b6d3a61"></a>🟥 2026-06-10
X·@scaling01 · 也是最好的多模态模型之一 · 模型发布
Fable 5 also seems to be one of the best, if not the best multimodal model
tags: 好观点 · → daily
📷 原图
- <a id="atom-62174d6b023c3340"></a>🟥 2026-06-10
X·@scaling01 · ALE-Bench 性能 · 模型发布/技术路线
Fable 5 absolutely crushing ALE-Bench
tags: 好观点 · → daily
value: qty=crushing · direction=正
📷 原图
- <a id="atom-0d52b457a39d9098"></a>🟥 2026-07-02
X·@ivanfioravanti · 性能状态 · 性能评价
still the ULTRA powerful model
tags: 好观点 · → daily
- <a id="atom-0219d4279718bbab"></a>🟥 2026-07-01
X·@Conor_D_Dart · AI游戏开发能力 · 产品发布
Fable 5 is a beast at ai game Dev!
tags: 好观点 · → daily
- <a id="atom-eabda0b634cca786"></a>🟥 2026-06-27
X·@AndrewCurran_ · approval for general release · 产品发布/监管政策
I think both Fable 5 and GPT-5.6 get approved for general release next week, and for use outside of the United States as well.
tags: 好观点 · → daily
value: date=next week
- <a id="atom-a392097afc595f76"></a>🟥 2026-06-27
X·@deredleritt3r · 通过测试后发布时间 · 模型发布
once this benchmark and 'rules of the road' are finalized, it will be announced that both GPT-5.6 and Fable 5 pass and can be released to the public.
tags: 好观点 · → daily
value: 事件=基准和规则确定后发布
- <a id="atom-32d7dfd4cb768894"></a>🟥 2026-06-22
X·@imjustnewatai · tier_status · 竞争格局
Fable 5 will still be the higher tier
tags: 好观点 · → daily
value: qty=higher tier · date=2026-06-22
- <a id="atom-6062fb6eb93da688"></a>🟥 2026-06-19
X·@Xianbao_QIAN · 被作为基座模型进行蒸馏 · 技术路线/模型发布
Very impressive local model 'distilled' on top of Fable 5!
tags: 好观点 · → daily
📷 原图
- <a id="atom-b2441930cfc53def"></a>🟥 2026-06-16
X·@thegoodreturn · intent inference performance · 模型发布
People loved Fable 5 because it inferred intent better than any other model by a considerable margin
tags: 好观点 · → daily
value: direction=better than any other model
- <a id="atom-d12ef5aa35da20ed"></a>🟥 2026-06-12
X·@MParakhin · is a step change in league of its own · 模型质量
I played with it extensively. I agree - it IS in a league of its own. A step change.
tags: 好观点 · → daily
- <a id="atom-72c65d78abeb1b9b"></a>🟥 2026-06-12
X·@htihle · reliability advantage over gpt 5.5 (xhigh) · 性能基准
the difference is mostly in reliability
tags: 好观点 · → daily
📷 原图
- <a id="atom-af39c8b43f6ff748"></a>🟥 2026-06-11
X·@teortaxesTex · ability vs GPT-5.5-Pro-Extended as computer scientist · 模型发布/技术路线
Fable 5 is indeed stronger than *GPT-5.5-Pro-Extended* as a *computer scientist*, even without a lot of thinking
tags: 好观点 · → daily
value: qty=stronger · direction=positive
- <a id="atom-3671fc6cc7d0de1b"></a>🟥 2026-06-10
X·@jerhadf · TTC scaling · 推理能力/技术路线
fable 5’s TTC scaling is incredible
tags: 好观点 · → daily
value: direction=incredible
📷 原图
🟥 空头 takes (bearish) (22)
- <a id="atom-bc802646265f87df"></a>🟥 2026-06-19
X·@dhtikna · 性能仅高出 3% · 模型发布/竞争格局
fable 5 only outperforms by 3%
tags: 好数字·好观点 · → daily
value: qty=3% · direction=higher
- <a id="atom-1b3d668f4d493f9b"></a>🟥 2026-06-17
X·@garrytan · 禁令导致的生产力损失 · 生产力损失
Rough estimate on $ productivity lost by Fable 5 ban: $12M per hour
tags: 好数字 · → daily
value: qty=$12M per hour · direction=loss
- <a id="atom-57b6f57fccb39913"></a>🟥 2026-06-09
X·@lu_sichu · safety filters not triggered despite harmful behavior · 安全事故
this behavior happens without the safety filters being triggered
tags: 好观点 · → daily
- <a id="atom-057cffe08f9eab3a"></a>🟥 2026-07-03
X·@bridgemindai · 每周使用限制 · 产品发布/竞争格局
No weekly limit that serious builders burn in 24 hours
tags: 好观点 · → daily
value: qty=24小时内被认真构建者用完
📷 原图
- <a id="atom-5afa1e7653f2c97d"></a>🟥 2026-07-02
X·@Hesamation · 性能评价 · 模型发布/技术路线
Fable 5 isn't nerfed, it's SLAUGHTERED.
tags: 好观点 · → daily
📷 原图
- <a id="atom-df5587ce3b8e87a8"></a>🟥 2026-07-02
X·@Hesamation · guadrails 限制 · 技术路线
the problem isn't even the model itself, but the hard guardrails Anthropic has set in place.
tags: 好观点 · → daily
📷 原图
- <a id="atom-430e223b90d8b628"></a>🟥 2026-07-02
X·@IlyasHairline · 安全漏洞修复能力 · 技术路线/安全事故
Fable 5 dodging every prompt from fixing code vulnerabilities
tags: 好观点 · → daily
- <a id="atom-87880fcdea191f20"></a>🟥 2026-07-02
X·@IlyasHairline · 气味识别能力 · 技术路线
Fable 5 dodging every prompt from fixing code vulnerabilities to the smell of bumblebee fart
tags: 好观点 · → daily
- <a id="atom-21e55b27916c47c3"></a>🟥 2026-07-01
X·@romilbijarnia · don't jailbreak · 安全事故
NOBODY FUCKING JAILBREAK IT NOW!!!
tags: 好观点 · → daily
- <a id="atom-c60da12e0569c418"></a>🟥 2026-07-01
X·@Web3Aible · 将无法正常使用，严重代码工程无法正常工作 · 产品发布
Fable 5 won't be usable at all, maybe Child prompting can But serious code engineering work won't work
tags: 好观点 · → daily
📷 原图
- <a id="atom-78f6c78665f0784f"></a>🟥 2026-07-01
X·@tinkerersanky · 底层模型实为 Sonnet 5 · 产品发布
It’s Sonnet 5 beneath it
tags: 好观点 · → daily
- <a id="atom-a5d9eb7bfbcf6f94"></a>🟥 2026-06-26
X·@AndrewCurran_ · expected region restriction · 模型发布
It would also almost guarantee that Fable 5 would get the same treatment.
tags: 好观点 · → daily
value: direction=US only
📷 原图
- <a id="atom-ab7dab8dc1e7fe1f"></a>🟥 2026-06-22
X·@imjustnewatai · comparison_to_GPT-5.6 · 竞争格局
still probably not fully Fable 5 level
tags: 好观点 · → daily
value: qty=higher tier · date=2026-06-22
- <a id="atom-dfba185822e56ad2"></a>🟥 2026-06-17
X·@rahulgs · will be the worst of this new class of models · 模型发布
Fable 5 will be the worst of this new class of models
tags: 好观点 · → daily
- <a id="atom-dd7bb84de0b338cb"></a>🟥 2026-06-13
X·@charles_maddock · 美国禁令后果 · 监管政策
Precedent for banning AI models from ie. @OpenAI, @Google, etc
tags: 好观点 · → daily
- <a id="atom-2201ea9b42ac53df"></a>🟥 2026-06-13
X·@yishan · contains cryptography considered dangerous · 技术路线/监管政策
it has a lot of cryptography which i understand to be dangerous
tags: 好观点 · → daily
- <a id="atom-bc59a35598cf2a1a"></a>🟥 2026-06-11
X·@tamaybes · code generation behavior · 模型发布
One interesting pattern with Fable 5 is that it will often say things that are gibberish when I use it for coding
tags: 好观点 · → daily
value: direction=negative
- <a id="atom-c158d1276a9d454f"></a>🟥 2026-06-11
X·@tamaybes · theory-of-mind failure · 技术路线
Fable explains that it invents codenames while reasoning about the problem, then fails to realize they're meaningless to me. Its neuralese is blending into its output because of a theory-of-mind failure about what's in its head vs. mine
tags: 好观点 · → daily
value: direction=negative
- <a id="atom-e241f92ec5807e7d"></a>🟥 2026-06-09
X·@rasbt · shadowbanning AI researchers · 监管政策
Fable 5 is shadowbanning AI researchers
tags: 好观点 · → daily
value: direction=active
- <a id="atom-69f27c98dd3f2dc0"></a>🟥 2026-06-09
X·@rasbt · ablation studies not visible to user · 监管政策
ablation studies are not visible to the user
tags: 好观点 · → daily
value: direction=not_visible
- <a id="atom-e0f25780f366058b"></a>🟥 2026-06-09
X·@brodieferguson · usage blocked flagged as security risk · 模型可用性
You can’t use Fable 5 for anything, because it’ll get flagged as a security risk.
tags: 好观点 · → daily
- <a id="atom-d5c5371da30f668b"></a>🟥 2026-06-09
X·@max_paperclips · score not representative due to fallback · 技术路线
so, you didn't actually score Fable then, you scored Opus.
tags: 好问题 · → daily
🟥 中性 takes (neutral) (12)
- <a id="atom-292b9b1c6e9b5294"></a>🟥 2026-06-28
X·@haivi_seo · 全球停用原因 · 出口管制
the global kill switch wasn’t because the model is magically too dangerous. it was because perfect nationality segmentation at frontier scale is currently impossible. export controls on weights + api access force a binary: geofence flawlessly or turn it off worldwide.
tags: 好观点·好思考 · → daily
value: direction=技术限制
- <a id="atom-f5917162de8c8ea8"></a>🟥 2026-06-25
X·@sammcallister · 可能为UI显示错误 · 产品发布
This could be a UI bug though, will track it down.
tags: 好思考 · → daily
📷 原图
- <a id="atom-73be9641ef4f32a4"></a>🟥 2026-06-22
X·@scaling01 · 对 Opus 4.7 性能对比 · 模型对比
it's not meaningfully better than Opus 4.7 and Opus 4.6
tags: 好观点 · → daily
value: direction=not meaningfully better than
📷 原图
- <a id="atom-37885e6984788a2f"></a>🟥 2026-06-23
X·@Yuchenj_UW · 延迟原因 · 监管政策
they don't want to be like Fable 5.
tags: 好思考 · → daily
value: qty= · date=
- <a id="atom-8eab5fa0a4adf863"></a>🟥 2026-07-02
X·@Hesamation · 路由问题 · 技术路线
this is a routing problem not the model itself.
tags: 好观点 · → daily
- <a id="atom-6f0fd2b6cd68a74d"></a>🟥 2026-06-22
X·@patience_cave · 发布预期 · 产品发布
no Fable 5
tags: 好观点 · → daily
value: qty=no · date=2026-06-22T17:27:50
- <a id="atom-bdb5be8f06d25632"></a>🟥 2026-06-15
X·@8teAPi · guardrail-bypass example basis for export controls · 技术路线/监管政策
this is kind of deeply silly
tags: 好观点 · → daily
- <a id="atom-d30ba4f5e9e59fef"></a>🟥 2026-06-13
X·@Alisvolatprop12 · 模型发布状态 · 模型发布
빠르면 일요일, 늦어도 다음주면 미국정부의 우려가 상당히 적용된 Fable 5의 재개
tags: 好观点 · → daily
value: direction=可能重启
- <a id="atom-8fd1ec71971e0b1c"></a>🟥 2026-06-12
X·@yishan · 出口管制禁令是否延续 · 出口管制
Maybe the Fable 5 ban will be resolved and it's all a big misunderstanding
tags: 好观点 · → daily
value: direction=uncertain
- <a id="atom-4b313b839ebbf741"></a>🟥 2026-06-11
X·@Alisvolatprop12 · 로컬 모델 출시 시기 · 产品发布
Fable 5급 로컬 모델은 올해 안으로 옵니다.
tags: 好观点 · → daily
value: date=올해 안
- <a id="atom-10bf86fc38b137f8"></a>🟥 2026-06-09
X·@Miles_Brundage · autonomous YouTube Poop generation task 搞笑程度 · 技术路线
Fable 5 is at least somewhat funnier than Opus 4.8 on the autonomous YouTube Poop generation task
tags: 好观点 · → daily
value: qty=at least somewhat funnier than Opus 4.8
- <a id="atom-4dfcea546096cd1a"></a>🟥 2026-06-09
X·@theo · daily use blocked only on Claude site · 模型可用性
I've been using it all day with no issues. Only got blocked on the Claude site
tags: 好观点 · → daily
⏱ 时间轴 (近 20)
- 🟦 2026-07-03 ·
fact · TerminalBench 表现 · → daily
- 🟥 2026-07-03 ·
fact · 每周使用限制 · → daily
- 🟥 2026-07-02 ·
narrative · 性能评价 · → daily
- 🟥 2026-07-02 ·
narrative · guadrails 限制 · → daily
- 🟥 2026-07-02 ·
fact · 路由问题 · → daily
- 🟥 2026-07-02 ·
narrative · 安全漏洞修复能力 · → daily
- 🟥 2026-07-02 ·
narrative · 气味识别能力 · → daily
- 🟥 2026-07-02 ·
narrative · SWE 基准测试提升 · → daily
- 🟥 2026-07-02 ·
position · 性能状态 · → daily
- 🟦 2026-07-02 ·
narrative · Web UI 使用限制 · → daily
- 🟦 2026-07-01 ·
fact · usage limit inclusion · → daily
- 🟦 2026-07-01 ·
narrative · 100% usage period actual · → daily
- 🟦 2026-07-01 ·
fact · 재출시 발표 · → daily
- 🟦 2026-07-01 ·
fact · available in subscription plan for 6 days · → daily
- 🟥 2026-07-01 ·
narrative · don't jailbreak · → daily
- 🟦 2026-07-01 ·
fact · 定价 · → daily
- 🟦 2026-07-01 ·
fact · 某些常规任务将回退至 Opus 4.8 · → daily
- 🟦 2026-07-01 ·
fact · 100% 使用率承诺期从至少两周缩减至 7 天 · → daily
- 🟦 2026-07-01 ·
fact · 新增 50% 每周使用限制 · → daily
- 🟥 2026-07-01 ·
forecast · 将无法正常使用，严重代码工程无法正常工作 · → daily
← 实体目录 · 系统日志
