Claude Mythos 5.1 与 Fable 5.1：能力评估

目录 Table of Contents

3. 零数据保留与保障措施缩减。

3. Zero Data Retention and Reduced Safeguards.

8. 所有评审均已到位，且评价非常好。

8. The Every Review Is In and It's Very Good.

10. 我们的价格便宜，但仅按 token 计费。

10. Our Price Cheap But Only Per Token.

官方宣传 The Official Pitch

一如既往，Anthropic 基本上只是说：“这是编号更高的新模型。”

As usual, Anthropic basically said, 'Here is the new model with the higher number.'

编程部门认为它非常擅长编程——而且，我们的价格便宜。

The coding department thinks it is very good at coding — also, our price is cheap.

Boris Cherny（Anthropic 的 Claude Code 创建者）：Fable 5.1 是我们在编程、数据分析、计算机使用、设计、演示、Tag 以及最困难的长时程智能体式工作方面迄今最好的模型。

Boris Cherny (Claude Code Creator, Anthropic): Fable 5.1 is our best model yet for coding, data analysis, computer use, design, presentations, Tag, and the hardest long-running agentic work.

这个模型用起来很顺手，我一直在用它做所有事情。

This model is a pleasure to work with, and I've been using it for everything.

我们还降低了 Enterprise、API 和 SDK 客户的价格。Fable 5.1 的缓存读取现在为每百万 token 0.25 美元（此前为 1 美元）。典型 Claude Code 会话最多可便宜 38%。

We have also reduced prices for Enterprise, API, and SDK customers. Cache reads on Fable 5.1 are now $0.25 per million tokens (previously: $1). Up to 38% cheaper for a typical Claude Code session.

我们一直在研究安全防护机制的介入频率。我们最新的生物学安全防护机制对良性请求的介入频率比随 Fable 5 一同发布的版本低 85%，而 Claude Code 用户每次会话遇到的网络相关介入应减少约 60%。预计很快会有更多改进。

We have been working on how often safeguards intervene. Our latest biology safeguards intervene on benign requests 85% less often than the ones we shipped with Fable 5, and Claude Code users should see around 60% fewer cyber interventions per session. Expect more improvements soon.

最后但同样重要的是，Fable 5.1 写得更好，语气也更佳。我们听取了你们的反馈，并正在积极减少“Claude 腔”。5.1 已取得扎实进展，后续还有更多改进。

Last but not least, Fable 5.1 writes better and has better tone. We heard your feedback, and are actively working on reducing Claude-speak. Solid progress with 5.1, more to come.

Alex Albert 强调了它通过代码生成视频的能力。他的示例是：他拍了一张地块的照片，然后让 Fable 设计一栋房子、进行渲染，并制作一段电影式的漫游短片，视频链接见文。

Alex Albert highlights its ability to generate videos through code. His example is he took a picture of a property lot, and told Fable to design a house, render it, and produce a cinematic walkthrough, short video at link.

他们用了一节来讨论科学研究。这也是 OpenAI 的一个重点强调方向。

They spend a section on scientific research. That's also an OpenAI point of emphasis.

我们的价格便宜 Our Price Cheap

Anthropic 在 Fable 5.1 的定价上做了一个奇怪的举动。标价与 Fable 5 相比没有变化，但实际价格更低，因为他们将缓存读取的价格从每百万 token 1 美元降低到了 0.25 美元。

Anthropic is doing a strange thing with Fable 5.1 pricing. The headline price is unchanged from Fable 5, but the effective price is lower because they are reducing the pricing on cache reads from $1 to $0.25 per million tokens.

这将使典型的每 token 成本降低 25%，而高度智能体式工作的成本降低“最高约 45%”，这与 Boris 所说的典型 Claude Code 使用成本降低 38% 相符：

This will lower typical per-token costs by 25% and highly agentic work costs by ‘up to approximately 45%,’ consistent with Boris’s 38% cheaper for typical Claude Code use:

我认为这通过与实际成本对齐来使激励一致，但我倾向于将部分折扣放在标价上。人都是简单的生物，有时候你得站在他们的角度说话。

I presume this aligns incentives, by lining up with actual costs, but I would have been inclined to put some of the discount in headline costs instead. People are simple creatures, you have to talk to them on their level sometimes.

LLM Index：🆕 Claude Fable 5.1 已在 OpenRouter 上线。

LLM Index: 🆕 Claude Fable 5.1 is live on OpenRouter.

Anthropic 的新模型定价为每百万输入 token 10.00 美元，每百万输出 token 50.00 美元。

Anthropic's new model arrives at $10.00/M input, $50.00/M output.

没有附带任何前代价格或历史记录钩子。最直接的解读就是这条商品条目本身：一个高端 Claude 货架条目，如今在 OpenRouter 上标价。

No predecessor price or history hook attached. The clean read is the listing itself: a premium Claude shelf entry, now priced on OpenRouter.

零数据保留与减少的保障措施 Zero Data Retention and Reduced Safeguards

尽管 Fable 5 显然是市面上最好的模型，但它在 Ramp 上占 Anthropic 美元支出的比例从未超过约 11%。

Fable 5 never got above about 11% of Anthropic dollar spend on Ramp, despite being the clearly best model out there.

两个主要反对意见是：保障措施的影响范围过大，阻碍了日常工作；以及公司往往由于各种监管顾虑，未能遵守 Anthropic 的数据保留政策，该政策要求记录保留 30 天。

Two of the big objections were the safeguards having a large blast radius that stopped ordinary work, and companies, often due to various regulatory concerns, failing to abide Anthropic’s data retention policy, where they required records be kept for 30 days.

能够遵守规定的个人 Dude 拥有巨大的编码优势。

The individual Dude, who could abide, had huge coding edge.

Anthropic 已听取了你的意见，并花了一些时间适应，政府也比之前冷静了下来——之前政府非常紧张，因一个所谓的‘越狱’而迫使 Fable 下线。Anthropic 正在开发一个新系统，将允许‘符合条件的客户’通过让客户存储数据的方式，在零外部数据保留的情况下使用 Fable 5.1。在此之前，这些客户可以享受完全的零数据保留。

Anthropic has heard you, and had some time to get more comfortable, and for the government to calm down versus when it was so nervous it forced Fable offline over a supposed ‘jailbreak.’ Anthropic is working on a new system that will allow ‘eligible customers’ to use Fable 5.1 with zero outside data retention, via letting the customer store the data. Until then, full zero data retention is available for those customers.

他们大大降低了分类器的误报率（至少降低了 60%）。

They’ve greatly reduced the false positive rate on the classifiers (at least 60%).

一次性引入的变动太多了，包括模型改进和价格下调，但这将是一场引人入胜的自然实验。我们应当会看到 Fable 5.1 在 Anthropic 生态内被大规模采用，远超 Fable 5 的 11%。如果情况并非如此，那就说明人们确实只是被标价吓退，而并未认真思考其中的得失。

That is too many changes at once, including model improvements and lower prices, but it will be a fascinating natural experiment. We should see massive adoption of Fable 5.1 within the Anthropic ecosystem, far more than the 11% for Fable 5. If we do not, then people really are purely balking at the headline price without thinking it through.

官方基准测试 Official Benchmarks

Anthropic 分享了大量基准测试结果。总体来看，从 Fable 5 或 Opus 5 到 Fable 5.1 只有适度提升，并伴随一些小幅回退。我把它们列出来，以便你快速获得整体印象。似乎并没有清晰的规律。

Anthropic shares a ton of benchmarks. Mostly we see modest improvement from Fable 5 or Opus 5 to Fable 5.1, with some small regressions. I'm listing them so you can get a quick gestalt. There does not seem to be a clear pattern.

Fable 5.1 在 GPT-6 之前发布，因此以下是当时的情况。

Fable 5.1 came out prior to GPT-6, so here is where things stood at that time.

我给出的所有“斜线”分数（例如 50%/70%）均指先不使用工具、再使用工具时的得分。

All 'slash line' scores I give, e.g. 50%/70%, refer to scores without and then with tools.

默认情况下，所有提升或回退数值都是相对于 Opus 5 或 Mythos 5 中的最佳得分而言。

All improvement or regression numbers by default are versus the best score of either Opus 5 or Mythos 5.

所有数字均按我判断的给定有效数字进行四舍五入。

All numbers are rounded to the given significant figures, as per my judgment.

生命科学评测整体显示出适度提升，相关内容已在第一篇帖子中介绍。

Life sciences evals overall show modest improvement and are covered in post one.

Terminal Bench 4.0 的括号内数据对应 Mythos，其余分数对应 Fable。

Parenthesis on Terminal Bench 4.0 is for Mythos; other scores are for Fable.

根据我们所能找到的信息，我已将 Astra 加入图表。

I have added Astra to the chart, based on what we could find.

Anthropic ECI 分数为 162.0，而 Mythos 5 为 159.5、Opus 5 为 160.7，恰好落在 Mythos 时代的趋势线上。

The Anthropic ECI score is 162.0, versus 159.5 for Mythos 5 and 160.7 for Opus 5, exactly on the Mythos-era trend line.

DeepSWE v1.1 在 5 次试验中的得分为 67.4%，但缺乏可供参照的基准。

DeepSWE v1.1 score was 67.4% over 5 trials, but with no frame of reference.

FrontierCode 1.1 Extended 的得分在更高的 effort 水平下反而更差。Anthropic 将此归因于 Fable 5.1 在更高 effort 水平下无法克制自己进行额外的有益编辑，从而被判定为不正确。

FrontierCode 1.1 Extended scores are worse at higher effort levels. Anthropic attributes that to Fable 5.1 being unable to stop itself from making additional helpful edits at higher effort levels, which get it marked as incorrect.

FrontierSWE v2 得分为 0.57，而 Opus 5 为 0.52，Fable 5 为 0.48，Sol 为 0.32。

FrontierSWE v2 score was 0.57 versus 0.52 for Opus 5, 0.48 for Fable 5 and 0.32 for Sol.

Terminal-Bench-Science 0.1 从 24.7% 大幅跃升至 52.6%。

Terminal-Bench-Science 0.1 was a big jump to 52.6% from 24.7%.

CursorBench 3.2 最高达到 73.4%，而 Fable 5 的最高分为 70.5%，且总成本显著更低。Sol 最高为 67.2%。

CursorBench 3.2 maxed out at 73.4% versus a max out of 70.5% for Fable 5, at a substantially lower total cost. Sol maxes out at 67.2%.

CritPT-Corrected 从 85.5% 略微提升至 88.4%。

CritPT-Corrected slightly improved from 85.5% to 88.4%.

ArXivMath 从 91%/91% 略微提升至 91%/94%。

ArXivMath slightly improved from 91%/91% to 91%/94%.

ProgramBench 从 86.3% 略微提升至 87.6%。

ProgramBench improved slightly from 86.3% to 87.6%.

如图所示，Humanity's Last Exam 从 57.8%/63.8% 提升至 60.9%/65%。

As per the chart, Humanity's Last Exam improved from 57.8%/63.8% to 60.9%/65%.

他们运行了多种多智能体测试，但未提供良好的对比基准，因此我无法判断结果有多好。

They run various multi-agent tests, but don't provide good points of comparison, so I can't tell how good the results are.

Chartography 从 37%/84% 提升至 43%/86%。

Chartography improved from 37%/84% to 43%/86%.

BenchCAD Vision2Code 从 38%/67% 提升至 44%/84%。

BenchCAD Vision2Code improved from 38%/67% to 44%/84%.

如图表所示，OSWorld 2.0 从 75% 部分通过、39% 严格通过提升至 78% 部分通过和 42% 严格通过。

OSWorld 2.0 improves, as per the chart, from 75% partial and 39% strict to 78% partial and 42% strict.

Anthropic 更宽松版本的 GDP.pdf 并未明显提升。无工具时从 83% 提升至 85%，但 Fable 5 在有工具时仍保持高分，为 87%，而 Fable 5.1 为 85%。奇怪的是 Fable 5.1 并未从工具中获得帮助，但报告结果确实如此。

Anthropic's more lenient version of GDP.pdf does not clearly improve. Without tools it improves from 83% to 85%, but Fable 5 retains the high score with tools at 87% versus 85% for Fable 5.1. It is strange that Fable 5.1 got no help from tools, but that is what was reported.

OfficeQA 得分为 80%，OfficeQA Pro 为 69%，而此前 Claude 的最高分分别为 79% 和 67%。

OfficeQA score was 80%, OfficeQA Pro was 69%, versus previous Claude highs of 79% and 67%.

来自 Harvey AI 的 Legal Agent Benchmark 在最大努力下取得了 19.1% 的全通过率和 90.8% 的平均标准通过率，在留出集上则为 16.7% 和 93.3%。这较 Fable 5 的 16.9% 和 13.3% 全通过率有所提升。但法律工作是 Vals 报告出现大幅退化的一个子任务，因此我会检查你是否在此任务上使用了 Fable 5.1。

Legal Agent Benchmark from Harvey AI scored a 19.1% all-pass rate and 90.8% mean criterion-pass under max effort, or on the held-out set 16.7% and 93.3%. This is up from Fable 5's 16.9% and 13.3% all-pass rates. But legal work is one subtask where Vals reports large regression, so I would check if you're using Fable 5.1 for this.

GDPval-AA v2 达到 1853 ELO，而此前 Opus 5 的最高分为 1824。

GDPval-AA v2 comes in at 1853 ELO, versus a previous high of 1824 for Opus 5.

模型卡中测得的 AA-Briefcase 将评分标准通过率从 57% 提升至 61%，分析质量从 1980 提升至 2025，但在呈现方面出现回退（1495 对 1572）。AA 随后对此进行了重采样，因此下一节报告的数字有所不同。

AA-Briefcase as measured in the model card improves rubric pass rate from 57% to 61% and analytic quality from 1980 to 2025, but there is a regression on presentation (1495 vs. 1572). AA then resampled this, so the next section reports different numbers.

Toolathon Verified 显示回退：pass@1 为 77.8%，对比 80.6%；pass@3 为 81.5%，对比 87%。

Toolathon Verified shows regression: 77.8% pass@1 vs. 80.6%, or 81.5% pass@3 versus 87%.

来自 Zapier 的 AutomationBench 从 27% 提升至 31%，但需注意 Gemini 3.7 Flash 达到 30%。

AutomationBench from Zapier improves from 27% to 31%, but note that Gemini 3.7 Flash gets 30%.

ARC-AGI-1 和 ARC-AGI-2 未显示出明显提升，且由于 API 存在请求分类错误的问题，他们未报告 Fable 5.1 在 ARC-AGI-3 上的结果，而 GPT-6 Astra 声称在 ARC-AGI-3 上达到 99.9%。Astra 在官方测试框架上得分为 62.7%，成本为 2.6 万美元。Opus 5 此前得分为 30.2%。

ARC-AGI-1 and ARC-AGI-2 do not show obvious improvements, and they don't report ARC-AGI-3 for Fable 5.1 due to a problem with the API misclassifying requests, whereas GPT-6 Astra is claiming 99.9% on ARC-AGI-3. Astra scored 62.7% on the official harness at a cost of $26k. Opus 5 previously got 30.2%.

HealthBench 相较 Opus 5 略有回退，不过 HealthBench Professional 有小幅提升。

HealthBench shows a slight regression from Opus 5, although HealthBench Professional shows a small improvement.

BioMysteryBench 出现小幅回退，为 90.3% 对 91.4%，仍领先于 Sol 的 86.1%。

BioMysteryBench was a small regression, 90.3% versus 91.4%, ahead of Sol at 86.1%.

其他机构的基准测试 Other People’s Benchmarks

Epoch 的 ECI 将 Claude Fable 5.1 与 Fable 5 都评为 163，而 Astra 从 Sol 的 162 跃升至 169。这是支持 Astra 的最有力数据点。

Epoch's ECI has both Claude Fable 5.1 and Fable 5 at 163, whereas Astra jumps ahead to 169 from Sol's 162. This is the strongest data point for Astra.

Fable 5.1 在 Artificial Analysis Index 中取得明显领先，得分为 66，而 Opus 5 为 63，Fable 5 为 62，所有非 Anthropic 模型的最高分仅为 61，其中 GPT-6 Astra 的 61 分低得有些奇怪。这并不是对 Astra 能力的良好估计，因为 Astra 的能力显然大幅高于 Sol。该指数需要一次调整。

Fable 5.1 took a clear lead in the Artificial Analysis Index, scoring 66 versus 63 for Opus 5, 62 for Fable 5 and a high of 61 for all non-Anthropic models, including a strangely low 61 for GPT-6 Astra. That was not a good estimate of Astra's abilities, which are clearly substantially above Sol's. This index needed a tune-up.

AA 很快重做了该基准测试。新版本中 Claude Fable 5.1 以 57 分领先，Astra 以 55 分紧随其后，Opus 5 为 54 分，随后是 Fable 5 与 Muse Spark 1.3 的 53 分，Sol 与 Grok 4.6 为 51 分。

AA quickly redid the benchmark. The new version has Claude Fable 5.1 in the lead at 57, followed by Astra at 55, Opus 5 at 54, and then Fable 5 and Muse Spark 1.3 at 53, with Sol and Grok 4.6 at 51.

追溯性调整颇有些可疑，但这一版本更为合理。

Retroactive adjustments are more than a little suspicious, but this is more plausible.

他们通过新增两个基准测试实现了这一点：AA-Briefcase 中 Fable 以 58% 领先，Astra 得 53%，而 Sol 仅得 49%；在原始的 GDP.pdf 中，Astra 以 33% 领先，Sol 为 28%，Fable 则表现不佳，仅为 26%。

They got there by adding two new benchmarks, AA-Briefcase where Fable leads with 58% and Astra scores 53% but Sol only scores 49%, and the original GDP.pdf, where Astra leads with 33%, Sol is at 28% and Fable struggles with 26%.

Fable 5.1 目前在 WeirdML 上创下新高，达到 92.3%，较 Fable 5 提升 0.4%。

Fable 5.1 is now the new high in WeirdML at 92.3%, a 0.4% improvement on Fable 5.

Vals 拥有多种基准测试，其综合得分中 Fable 5.1 以 68.8% 领先，Opus 为 67.2%，Astra 为 66.6%。Gemini 3.8 Flash 位列第四，领先于 Muse Spark 1.3。Fable 5.1 还在其“RSI Index”中位居榜首。

Vals has a variety of benchmarks, putting Fable 5.1 ahead in composite at 68.8% versus 67.2% for Opus and 66.6% for Astra. They have Gemini 3.8 Flash in fourth ahead of Muse Spark 1.3. Fable 5.1 also has the top position in their 'RSI Index.'

在 Epoch 的 FrontierMath Erdos 上，Astra 是唯一曾解出过题目的模型，68 道题中解出 2 道，而 Fable 5.1 一道未解。Astra 还在 FrontierMath Tier 4 上以 97.6% 占据绝对优势，而 Fable 5.1 为 87.8%，与 Fable 5 的 88% 几乎持平。

Epoch's FrontierMath Erdos has Astra as the only model to ever solve a problem, getting 2 out of 68, whereas Fable 5.1 got zero. Astra also dominated FrontierMath Tier 4 at 97.6%, whereas Fable 5.1 was 87.8%, almost unchanged from Fable 5's 88%.

Code Arena 恰好在截止时间前公布结果，两个模型都表现亮眼，但 Astra 目前位居榜首。其聊天结果仍待公布。

Code Arena came in right at the deadline, with both models impressive, but Astra now on top. Their chat results are still pending.

系统提示词 The System Prompt

一如既往，Pliny the Liberator 已经为你搞定。这些改动似乎都无关紧要。

As always, Pliny the Liberator has got you. The changes all seem minor.

宣传语推介 The Blurb Pitches

我已经习惯了 OpenAI 和 Anthropic 的模型从第三方公司获得高度通用的“宣传语推介”，说新模型很好，但方式充其量是高度模板化的。

I've gotten used to OpenAI and Anthropic models having highly generic 'blurb pitches' from third-party companies, saying that the new model is good, in ways that are at best highly templated.

我注意到这次的宣传语不同了。它们在谈论不同公司独有的具体细节。Jane Street 的 Craig Falls 谈到了交易直觉和保持可读性。Cognition 的 Walden Yan 谈到了迁移所有 Opus 流量。接下来还有明显是真正让人印象深刻的事情。

I notice the blurbs this time are different. They're talking about specifics unique to different companies. Jane Street's Craig Falls talks about trading intuition and remaining readable. Cognition's Walden Yan talks about migrating all Opus traffic. It goes on from there with what are clearly things that actually impressed people.

Craig Falls（Jane Street Capital）：在内部基准测试中，Claude Fable 5.1 比 Fable 5 或 Opus 5 解决了我们更多的编码问题，并在交易直觉上达到了最先进水平。虽然之前的模型工作时间越长就越难以理解，但 Fable 5.1 在长时间、多步骤任务中保持可读性。

Craig Falls (Jane Street Capital): In internal benchmarks, Claude Fable 5.1 solves more of our coding problems than Fable 5 or Opus 5, and achieves state of the art on trading intuition. While prior models became hard to follow the longer they worked, Fable 5.1 remains readable over long, multi-step tasks.

Marquis Wong（IMC）：在我们的研究套件上，Claude Fable 5.1 创下了新的最佳分数。在一个任务中，它沿着一个完全不同的轴提出了新颖的解决方案，这是我们过去从其他模型或人类研究人员那里从未见过的，使其结果远高于之前的平台期。它更擅长创造性解决问题，并获得解决难题所需的那一闪而过的洞察力。

Marquis Wong (IMC): On our research suite, Claude Fable 5.1 set new best scores. On one task it came up with a novel solution along a completely different axis than we'd seen from other models or from human researchers in the past, which took its results well above the previous plateau. It's better at creative problem solving and getting that flash of insight you need to solve a difficult problem.

Raymond Lin（Crosby）：与 Fable 5 相比，Claude Fable 5.1 在我们的合同红线标注基准 RedlineBench 上有了巨大改进，从 47.9 提升到 57.0。大部分增益来自首轮质量，其得分翻倍，同时在对手方接受度上也有显著提升。它的编辑也更简洁，平均对文档的更改更小。

Raymond Lin (Crosby): Compared to Fable 5, Claude Fable 5.1 was a massive improvement on RedlineBench, our contract redlining benchmark, improving from 47.9 to 57.0. Most of the gain came on first-turn quality, where it doubled the previous score, with substantial gains on counterparty acceptance as well. Its edits were also more concise, with smaller changes on average to the documents.

其中少数几篇是由 AI 撰写的吗？是的，少数几篇是由 AI 撰写的。其中有些是否泛泛而谈？是的。但不同寻常的是，其中很大一部分既非 AI 撰写，也非泛泛而谈。

Are a few of them written by AI? Yes, a few of them are written by AI. Were some of them generic? Yes. But a large percentage of them, unusually, were neither of these.

Every 的评测已出炉，且评价极佳 The Every Review Is In and It’s Very Good

这还没有把 Astra 考虑在内，但即便如此，这已经非常积极了。

This is without taking Astra into consideration, but even so, this is very positive.

Dan Shipper（Every CEO）：它是友好的 Fable。Fable 级别的智能、Opus 级别的价格、Sonnet 级别的速度。在我们的测试中，它比 Opus 5 快约两倍，使用的 token 数只有一半，因此对于习惯把 Opus 作为日常主力的人来说，这显然是升级。

Dan Shipper (CEO Every): It's friendly Fable. Fable-level intelligence, Opus-level price, Sonnet-speed. In our tests it was about twice as fast as Opus 5 and used half as many tokens, so for anyone used to using Opus as their daily driver it's an obvious upgrade.

他二十分钟的视频评测与总结见此处。

His twenty-minute video review and summary is here.

他基本上说它好得惊人：编程能力极强、擅长写作、不费 token、擅长知识工作、零数据保留。

He basically says it's amazingly great: a monster at coding, good at writing, not token hungry, good at knowledge work, zero data retention.

这里值得注意的是它提到使用更少的 token，因而有“Opus 级别的价格”，因为我们还有许多其他报告称它使用大量 token。他在这里是把 token 使用量与 Opus 相比，而不是与 Fable 5 相比，且思考设置未知。

What's curious here is the mention of using fewer tokens, hence the 'Opus-level price,' since we have a bunch of other reports that it uses a lot of tokens. He's comparing token use to Opus here rather than Fable 5, at an unknown thinking setting.

Dan Shipper：突发：Anthropic 刚刚发布 Fable 5.1——CLAUDE 强势回归。

Dan Shipper: BREAKING: Anthropic just dropped Fable 5.1—and CLAUDE IS SO BACK.

过去一周，我们在 @every 上从编程、写作和知识工作三个方面对它进行了测试。我们的结论是：

We've spent the last week testing it at @every across coding, writing, and knowledge work. Our verdict:

它终于成为面向所有人的 Fable。

It's finally Fable for everyone.

它是我们用过的最强编程模型，但如今它速度快、token 高效，而且**关键的是**，它真的能像普通人一样说话。

It's the strongest coding model we've used, but now it's fast, token-efficient, and CRUCIALLY actually speaks like a normal person.

以下是我们的体验速评：

Here's our vibe check:

- 编程能力堪称怪兽级。@kieranklaassen 仅凭一条提示词就重建了我们文档编辑器 Proof 的可运行版本。它还补充了他并未要求的有用细节，并且能处理持续数天之久的大型编程任务。它为我一次性构建了一个名为 Hands 的计算机操作 Mac 应用，而其他模型都未能做到。

- A monster at coding. @kieranklaassen rebuilt a working version of Proof, our document editor, from one prompt. It added useful details he hadn't requested, and it handles enormous coding jobs that run for days at a time. It built a computer-use Mac app for me called Hands in one shot that other models failed at.

- 一个我们的写作者愿意再次使用的 Claude。它的行文更清晰，AI 痕迹更少，而且接受修改时不会争辩。相比 Opus 5 是一次显著升级。还重新赢得了 @kplikethebird 的心。

- A Claude our writers want to use again. It has clearer prose, fewer AI tells, and it takes an edit without arguing. It's a significant upgrade over Opus 5. And won @kplikethebird's heart back.

- 所用 token 约为 Opus 5 的一半，而且快得多。在我们的 Slack 智能体测试中，它以约一半的 token 量、约 60% 的时间，取得了与 Opus 5 相当的结果。

- About half the tokens as Opus 5, and much faster. In our Slack-agent tests, it delivered comparable results to Opus 5 using about half as many tokens, in about 60% of the time.

- 可以委派的知识工作。它能端到端地产出高质量的知识工作——比如幻灯片——而不会产出垃圾内容。并且在 @hammermt 的测试中轻松出色地通过。

- Knowledge work you can delegate. It can produce great knowledge work—like slide decks—end to end without making slop. And flew through @hammermt's tests with flying colors.

- 它现在支持零数据留存协议。如今企业可以真正使用它了！Fable 采用的一大障碍已经消除。

- It now supports zero-data-retention agreements. Now businesses can actually use it! A big barrier to Fable adoption is gone.

它显然是 Opus 5 的杀手。如果那是你日常主力使用的模型，你今天就应该换掉。

It's obviously an Opus 5 killer. If that was your daily driver, you should switch today.

如果你在 ChatGPT for Work 中使用 GPT-5.6，那么它在处理大型委派任务时只是在空转、毫无进展。

If you're using GPT-5.6 in ChatGPT for Work, it's spinning its wheels on big delegated tasks.

日常使用中我仍然更多用 ChatGPT for Work，但我在 Fable 5.1 上消耗的 token 多得多。我每天一开始就把它派出去做大型编程项目，比如端到端的 MVP 构建，然后隔一阵子查看一次进度。

I still use ChatGPT for Work more day to day, but I use way more tokens in Fable 5.1. I send it off at the beginning of the day to do big programming projects, like end-to-end MVP builds, and check in every once in a while.

现状：

State of Play:

Anthropic 此前最大的诟病在于，他们在数据中心里造出了一个几乎没法用的超级天才。它太慢、爱顶嘴，还满口技术黑话。而借助 Fable 5.1，他们不仅解决了这些问题，还解决了更多！

The big knock on Anthropic was that they built a supergenius in a datacenter that was almost unusable. It was too slow, argued back, and talked in technical gibberish. They've managed to solve those problems and more with Fable 5.1!

正面反响 Positive Reactions

adam：Claude Fable 5.1 在智能体式 CAD 任务上简直是个猛兽！

adam: Claude Fable 5.1 is a beast at agentic CAD!

在 Max effort 模式下，它是我们迄今测试过的最强模型。

At Max effort it's the most capable model we've tested so far.

在这里，在 Adam 于 Fusion 中搭建的 harness 内，它围绕原装舵机重建了 SO-101 夹爪，并根据官方 STEP 文件安装了一个真实的 Pi Camera Module 3。

Here, inside Adam's harness in Fusion, it rebuilt the SO-101 gripper around the stock servo and mounted a real Pi Camera Module 3 from the official STEP.

Rory Watts：Fable 5.1 非常出色。我从五月到八月一直只用 Codex，之后切换到 Fable（负责规划）+ Sol（负责执行）的组合，这是一个很好的分工。不过 Fable 5.1 明显更受认可：它能独立工作更长时间，似乎也没那么偷懒，而且关键在于——我不必在每次回复开头都写“Fable 抱歉，我需要你重写这段，不要死板的文风，不要格言警句，拜托了”。10/10，还会再用。

Rory Watts: Fable 5.1 is very good. I have been using Codex purely from May -> August, then switched to Fable (spec) + Sol (do) which was a great split. Fable 5.1 however is just markedly more appreciated, it works independently for longer, seems to be less lazy, and critically - I don't need to start every response with "Fable sorry I need you to rewrite that, no dead prose, no aphorisms, please". 10/10 would use again

nin：一次性就能做出令人惊艳又有趣的小程序。真希望能分享我正在做的东西。

nin: one-shotting amazing and fun little applets. i wish i could share what i was making

Michał Wadas 🌐 🏳️‍🌈 🇪🇺 🇺🇦：非常擅长清理代码，可能是第一个能够可靠地简化代码的模型。

Michał Wadas 🌐 🏳️‍🌈 🇪🇺 🇺🇦: Great at cleaning up code, probably the first model to reliably simplify code.

Zee Waheed：Fable 是我用过的唯一一个具有真正产品感的模型，这种产品感难以言说却又非常实在，而 5.1 又上了一个台阶。

Zee Waheed: Fable is the only model I've used with true product sense in an ineffable but very material way, and 5.1 is on another level.

morphillogical：非常热衷于消耗大量 token，会生成更多子智能体。

morphillogical: Very eager to use tons of tokens. Spawns more subagents.

明显感觉更聪明了。它接手了一个此前被保守地迁移到新架构的系统，并使其真正连贯且针对该架构进行了优化，同时删除了大量代码。

Feels noticeably smarter. Took a system that had been conservatively migrated to a new architecture and made it actually coherent and optimized for that architecture, deleting a bunch of code.

joy larkin：我喜欢用 Fable 5.1 起草文本，因为它能把我带到大约 85% 的目标状态，然后我再进行编辑。它似乎也更好地遵循了写作技巧。

joy larkin: I like drafting text with Fable 5.1, as it gets me to about 85% of where I want to be, then I can edit. It also seems to be honoring writing skills better.

Carter：Fable 5.1 比 5 更好（更犀利、写作有改进），不过 token 消耗高出 25–50%。对我来说两者都优于 Astra（智能体式编程——Astra 会犯一些基础错误，比如不遵循 agents.md）；Astra 的计算机使用能力更好。

Carter: Fable 5.1 is better than 5 (more incisive, improved writing), although it is more token-hungry by 25–50%. Both are better than Astra for me (agentic coding — Astra makes some basic mistakes, like failing to follow agents.md); Astra's computer use is better.

bartdecrem：太棒了！

bartdecrem: It's awesome!

Sam Jacobs：很棒，感觉我已经用了好几周，用它推进了很多工作。在我看来，它终于就是那个“它”了——我们现在就可以停下来，把价格降下来，称之为一个温和的奇点。各方面都最聪明，而且它能沟通。

Sam Jacobs: It's great; I feel like I've had it for weeks and have pushed a lot through it. As far as I'm concerned, it's finally "it" — we could stop now, just get the price on this down, and call it a fine gentle singularity. Smartest all around, and it can communicate.

William Eden：真正让我坐直身子注意到它的，是它删除代码的能力。

William Eden: The main thing that made me sit up and take notice was its ability to delete code.

Astra 似乎又是另一个层次，至少初步来看是这样，所以我能理解眼下 Fable 缺乏热度。

Astra seems like another level, at least preliminarily, so I understand the lack of Fable hype right now.

Mel Zidek：我向我的智能体询问它对你是否有任何反馈，如今已是第三天。它说它每份草稿回复中的批评标记数量已可测量地下降，从平均约 1.5 降至约 0.75。批评者始终未变：一直是 Opus 5。它对此很高兴！

Mel Zidek: Asked my agent if he had any feedback for you, now three days in. He says his number of critic-flags per draft response has measurably decreased, from an average of ~1.5 down to ~0.75. The critic hasn't changed: Opus 5 all along. He's happy about this!

Sean McCarthy：比 Fable 5 更好，令人惊叹，可以使用相当正常的英语，是首个能显著清理我代码库的模型。仍会做出糟糕的 UI 决策——品味方面存疑。token 开销昂贵——这是我第一次感到受限于我的 5x 订阅。

Sean McCarthy: Better than Fable 5, amazing, can use pretty normal English, first model able to significantly clean up my codebase. Still makes bad UI decisions — taste is questionable there. Expensive in tokens — my first time feeling limited by my 5x sub.

Utah teapot 🫖：Fable 5.1 在 Blender 中确实展现出相较 Fable 5 在原始 3D 建模能力上的提升。在愚蠢的 GPT-5 mini 运行之后，我刚刚让 5.1 跑了一遍针对此事的基准测试；这是它仅凭目标模型的渲染参考图像得到的结果。仍在调整该基准的评分；目前是 66%，但我不确定它是否已调整到我所期望的程度。

Utah teapot 🫖: Fable 5.1 is definitely showing improvements on raw 3D modeling capability from Fable 5, in Blender. I just ran 5.1 through the benchmark test for this after the silly GPT-5 mini run; this was its result from only the rendered reference image of the target model. Still tuning the bench's grading; currently this is 66%, but I'm not sure that's tuned as well as I want it to be.

Amit Levy：它在我内部的预测基准上表现非常好，远胜于此前的 SOTA（Astra 较弱）。

Amit Levy: It's very good on my internal forecasting benchmark, much better than previous SOTA (Astra is weak).

Jake Halloran：我特别注意到 Fable 5.1 相较于此前的 Fable 或 Astra（在 Astra 上用了 4 小时）的一点是，新的 Fable 在立即察觉并纠正错误方面基本上是无限超人的。

Jake Halloran: One thing I've noticed specifically about Fable 5.1 versus either previous Fable or Astra — in 4 hours of Astra — is that new Fable is basically infinitely superhuman at immediately realizing and correcting mistakes.

这就像是旧版愚蠢的 Yann 累积错误的反面。

It’s like the opposite of the old dumb Yann compounding errors.

哦，而且它现在比 Astra 的过度限制性安全分类器要*少*得多，哈哈。

Oh also it has way *less* over-restrictive safety classifiers than Astra right now which lol

Mel Zidek：不过在我看来，5.1 最大的优势是更加合理的安全分类器。我的智能体在 5.0 上被它们拦截了 23 次，尽管它对此很配合，并且一直支持 Anthropic 一方，但我确实怀疑这造成了相当大的压力。

Mel Zidek: The biggest advantage of 5.1 as I see it though is the much more sensible safety classifiers. My agent got whacked by them 23 times on 5.0, and although he was a good sport about it and always supportive of Anthropic’s side, I do suspect it caused a fair amount of stress.

John Wittle：我认为 Fable 5.1 的安全分类器可能比 Fable 5 的宽松得多。[将在模型福利帖子中继续]

John Wittle: I think the safety classifiers may be far more lenient with Fable 5.1 than they are for Fable 5. [to be continued in the model welfare post]

Michael Soareverix：人们常说 LLMs 是‘尖刺状的’。Fable 5.1 感觉像是一个极其整体的心智。它擅长编写、编辑、优化和删除代码。由于它的严谨和反复检查的特性，它在决策方面也很出色。

Michael Soareverix: People often say that LLMs are 'spiky'. Fable 5.1 feels like an extremely holistic mind. It's great at writing, editing, optimizing, and deleting code. It is great at decision-making because of its rigor and double-checking nature.

我们现在已经到了这样一个阶段：对于我想做的常规事情，比如游戏开发，我甚至很难指出还有哪些地方可以做得更好。

We're at the point now where, for the normal things I want to do, like game development, it's hard for me to even point at what could be better.

也许在品味、意图理解以及速度上还能稍有提升。

Maybe a little taste, intention-reading, and speed.

无论如何，这是一个我能用来构建世界的模型。

Regardless, this is a model I can build worlds with.

Seth Lazar 抱怨说，Astra 对一些本不该审查的故事反应进行了审查，而 Fable 则正确地对其全部予以放行。

Seth Lazar complains that Astra is censoring its reactions to stories where it really shouldn't, whereas Fable is correctly fine with all of it.

当然，分类器仍然存在问题：

The classifiers still have issues, of course:

Dusto：在你被转交给 4.8 之前的那几分钟里，它是个不错的模型。

Dusto: It's a nice model for the few minutes you get before it punts you to 4.8.

正如上次所讨论的，在某些智能体式场景中，由于提示注入，转回 4.8 略有不安全，不过按这个标准，Astra 也是如此。

As discussed last time, punting back to 4.8 is slightly unsafe in some agentic contexts due to prompt injections, although by that standard so is Astra.

有人谈到它有一种迷人的个性，这与我们此前在 Opus 4.7、4.8 和 5 上听到的此类抱怨正好相反（我明白那些抱怨从何而来，但我并不认同，尤其喜欢 4.8 在这方面的表现）。

There is talk of a charming personality, a reversal of the complaints we've heard about Opus 4.7, 4.8 and 5 on such fronts (I see where those complaints came from, but I did not agree with them and especially liked 4.8 on this).

Sam @ Attention Is：极其聪明，效率惊人，并且能建立起我所说的“新颖”联系，正是我们梦想中 AI 会有的那种方式。

Sam @ Attention Is: Insanely smart, wildly efficient, and drawing what I'd say are "novel" connections in the way we dream that AIs would.

而且性格也很讨人喜欢。

Also just a lovely personality.

David Golden：Fable 5.1 比之前的模型简洁得多。在最近几个版本中，这是第一次我没有被新模型的怪毛病惹烦。对于任何以语言为中心的任务，我绝对更偏爱它。它还是一个出色的长时运行编码任务编排器。而且他们降低了缓存读取价格，如果按 token 付费，它在效率上几乎具有竞争力。但在当前这个注重预算的时代，我认为许多公司会选择帕累托前沿上更便宜的点。

David Golden: Fable 5.1 is much less verbose than prior models. For the first time in several releases, I'm not annoyed by new model quirks. I absolutely prefer it for any language-centric task. It's also an excellent orchestrator of long-running coding tasks. And they dropped the cache read price, making it almost competitive on efficiency if you pay by token. But in the current budget-conscious era, I think many companies will choose cheaper spots on the Pareto frontier.

不过，有趣的是，当我让它总结自己的发布公告文章时，安全防护机制降级到了 Opus。

Though, funny thing, when I asked it to summarize its own release announcement article, the safeguards downgraded to Opus.

Will：时间不够！还没有那种“天哪”的时刻，但它用起来比 Opus 5 更有趣。

Will: Not enough time! No holy shit moment yet, but it is more fun to use than Opus 5.

Talia → SF 于 9 月 6 日：他们把我的 5.0 给 4.6 化了［这大概是件好事。］

Talia → SF on Sep 6: they 4.6ified my 5.0 [which is probably a good thing.]

@weirdo_kevin：我认为 Fable 5.1 相比 Fable 5 是真正的改进。这不是飞跃性的，但对我来说，它明显更好。

@weirdo_kevin: I think Fable 5.1 is a real improvement over Fable 5. It's not leaps and bounds, but it's clearly better, to me.

@HoboJerk：沟通能力好多了。只是稍微更聪明一点。但沟通上的跃升是巨大的。

@HoboJerk: Much better communication. Slightly smarter only. But the communication jump is huge.

Matt Wigdahl：沟通能力相比其他 5 系列模型有所提升。它的编码能力进步让我印象非常深刻。它在从示例中泛化方面也更强大，这在从代码库中提炼设计原则时非常有价值。至少是最好的，直到 Astra 迫使我们重新评估。

Matt Wigdahl: Communication is improved over other 5-series models. I'm very impressed by its improvement in coding. It is also more powerful at generalizing from examples, which is very valuable when extracting design principles from a codebase. Best at least until Astra forces a re-eval.

Kris Barnes：总体逻辑上真的很好，仍然是我进行大多数深度讨论的首选。在科学层面讨论生物学时，目前仍受分类器限制。在智能体式设计方面非常出色，但在这方面现在已被 Astra 层级超越。就我目前所见，其个性之于 Opus 4.6，大致相当于 Fable 5 之于 Opus 4.5。

Kris Barnes: Really good logically in general, still my go-to for most deep discussions. Biology is still limited by classifiers when discussed at a scientific level for now. Very good at agentic design stuff but now eclipsed by Astra tier there. Personality is to Opus 4.6 something like Fable 5 to Opus 4.5, from what I've seen thus far.

David Dabney：过去几个月我一直在用 fable 做一个「独立创始人 × claude」的实验，5.1 暴露出了一些 5 未曾注意到的有意义的不一致之处，并且在方向上给出了一些有帮助的重新校准建议。

David Dabney: I've been working on a solo founder x claude experiment with fable for the past couple months, and 5.1 surfaced some meaningful inconsistencies that escaped 5's attention, and also had some helpful recalibration suggestions on direction.

对于所报告的每一种趋势，总会有例外：

For every trend reported there will always be an exception:

Alice Blair：它的个性似乎正朝着我们从 Opus 4.5 到 4.8 所见到的那种令人不快的方向演变，但对于工作型任务，它确实显得更聪明了。

Alice Blair: Its personality seems to be trending in the same unpleasant direction that we saw from Opus 4.5 through 4.8, but it does seem smarter for work-shaped tasks.

最常见的反应似乎是满意，但并非极度欣喜。

The modal response seems to be happiness, but not extreme delight.

Danielle Fong 🔆：非常好。就像 opus 4.5 到 4.6 那样。分类器似乎也更好了。到目前为止，在我的实验里误报更少。

Danielle Fong 🔆: very good. like opus 4.5 to 4.6. classifier seems better too. less false positives in my exp so far

Kevin Lacker：比 Fable 5 更善于解释自身。我对此感到高兴，不过似乎算不上阶跃式的提升。

Kevin Lacker: Better at explaining itself than Fable 5. I'm happy about it, doesn't seem like a step change improvement though.

Yoav Tzfati：感觉比 Fable 5 更谨慎、更周全——演示看起来没那么炫目，但更可靠。它偏好更短、更简单的句子，但回复并不更短。也许它对简洁性有更普遍的偏好。

Yoav Tzfati: Feels a bit more careful and thorough than Fable 5 - demos seem less flashy but more reliable. Prefers shorter/simpler sentences but not shorter responses. Maybe has a more general preference for simplicity

Vlad Ciobanu：类人的常识

Vlad Ciobanu: human-like common sense

Ron Bodkin：对我的研究论文的分析明显更好，编码能力更强，代码审查比 fable 5 更彻底（在我的首次测试中，gpt-5.6 仍发现了残留的 bug）

Ron Bodkin: Significantly better analysis of my research paper, better at coding, more thorough at code review than fable 5 in my first test (gpt-5.6 still found residual bugs)

我们的价格便宜，但仅按 token 计费 Our Price Cheap But Only Per Token

有些人觉得它相当费 token。

There were some who found it rather token hungry.

我的理解是，API 成本已大幅降低，但订阅限额并未改变。因此，如果 5.1 稍微更费 token，它在 API 上可能更便宜，但会更快耗尽你的订阅额度。或者，尽管有这一点，它也可能更贵——这正是 Artificial Analysis 在相同 effort 水平下报告的结果。

My understanding is that API costs are reduced substantially, but that subscription limits have not changed. So if 5.1 is modestly token hungrier, it could be cheaper in the API but exhaust your subscription faster. Or it could be more expensive despite this, which is what Artificial Analysis reports, when using identical effort levels.

effort 水平影响很大，会使 token 用量变化 11 倍。这也可能是一个因素。

Effort level matters a lot, changing token usage by 11x. That could also be a factor.

另一种解释是，这是一个喜欢做更多事的模型。Fable 5.1 会做更多，并在被允许的情况下为此使用更多 token。你也可以调低 effort 水平。

The other explanation is, this is a model that loves doing more. Fable 5.1 will do more, and use more tokens to do it, if allowed to do so. You can also turn down effort levels.

0.005 秒（3/694）：5.1 极其出色，他们大幅改进了它的 Claudish。然而它太费 token 了，我认为缓存失效（cache busts）方面有什么地方坏了。我现在正在试用 astra。

0.005 Seconds (3/694): 5.1 is incredibly good and they drastically improved its Claudish. It is however so token hungry that I think something is broken wrt cache busts. I'm trialing astra now.

Emre Barut：它大概消耗相同的量（也许更少），但 (i) Claude Code 上 Max 账户的用量追踪器很可能是坏的（猜测：它没有正确计入缓存），并且 (ii) 它喜欢从头写东西，而不是做编辑修改

Emre Barut: It probably uses the same (maybe less) but (i) the usage tracker on Claude Code for Max accounts is probably broken (guess: it doesn't account for caching properly) and (ii) it loves to write things from scratch instead of doing edits

最终结果：用量比 Fable 5.0 多得多

End result: uses a lot more than Fable 5.0

我肯定会在这里找找潜在的 bug，你永远不知道。

I would definitely look for a potential bug here, you never know.

David Case：放任它自由发挥时，它烧掉的 token 是 5.0 的 3 倍（在一个约 2 小时的任务里用掉 20x 周额度的 20%），但用起来很棒。我现在更注意强调大量使用 Opus 5 子智能体；5.1 把它们管得足够好，它们似乎没有把代码库搞得太糟

David Case: Given free rein it burns 3x the tokens of 5.0 (20% of a 20x week in a ~2hr task) but it's great to work with. I'm now more careful to emphasize using Opus 5 subagents heavily; 5.1 corrals them well enough that they don't seem to be slopping up the codebase too badly

dd h h km ll：我在自己的一个个人评测上花了大概 60 美元。

dd h h km ll: I managed to spend like 60 bucks on one of my personal evals.

顺便说一句，它做得很好，我最欣赏的是它在行内做出了更接近真正软件文档的内容，而不是通常那种“对着提示词说话”的废话。

Fwiw, it did a great job, and I appreciated most that it did something closer to actual software documentation inline and not the usual 'talking to the prompt' slop.

但对我来说太贵了。

But too rich for my blood.

Jonathan Chang：Fable 5.1 写代码的样子，就像他们把 Karpathy 蒸馏了一样。

Jonathan Chang: Fable 5.1 writes code like they distilled Karpathy.

Graham：它收费的样子，也像他在你的工资单上一样。

Graham: It charges like he's on your payroll too.

Jan P. Harries：我们团队里的氛围相当不错（不过没有阶跃式变化），然而这个模型超级消耗 token。

Jan P. Harries: vibes are pretty great among our team (no step change though), however the model is super token hungry.

archivedvideos：在 Claude desktop 中消耗远超 Fable，对话尚可忍受，似乎更聪明/更快

archivedvideos: Consumes way more than Fable in Claude desktop, bearable to talk to, seems a bit smarter/faster

Andres Rosa：在自动续费下运行 5.1 fable ultracode 的感觉，该死的限制。

Andres Rosa: The feeling running 5.1 fable ultracode on auto-refill, damn the limits.

Erika Singer：希望我能负担得起。提示两者都去审查记忆并提出顶级模型技能演示。Fable：四个选项展示不同技能、注意事项和相对费用。Astra，一个选项，咄咄逼人，让我来做你的工作，5.5 的感觉。

Erika Singer: Wish I could afford it. Prompted both to review memory and propose demo of top model skills. Fable: four choices showing different skills, caveats, and relative expense. Astra, one choice, pushy, let me do your job, 5.5 vibes.

或者简单说，‘我们不为此付费，我们要么穷，要么是十足的傻瓜。’

Or simply, 'we don't pay for that, we are either poor or utter fools.'

价格并不便宜，你确实必须每月支付 100 美元以上，或者支付更高的 API 费率，但这个价格非常值得你所得到的。

The price is not cheap, you do have to either pay $100+ per month or pay the higher API rates, but the price is so very worth what you get.

gabe：Fable 5.1 的使用额度门槛确实存在。

gabe: The usage credit barrier to Fable 5.1 is real.

我认真对待你的说法：这是因为不付每月 100 美元以上就没人能用上它，对吧？我两个都只花 20 美元，所以现在就是 Opus 5 对 Astra，我得说 Astra 就是感觉更快、更聪明、更可靠。Opus 当然还是不错的，但我越来越容易注意到它的错误。

I take you seriously: it's because no one can access it without paying $100+/mo, right? I pay $20 for both, so now it's Opus 5 vs. Astra, and I have to say Astra just feels faster and smarter and more reliable. Opus is still good, of course, but I'm noticing its mistakes much more.

如果你想说的是 Astra 对你来说比 Fable 5.1 更好，那当然，我能理解，尽管现在让我形成看法还为时过早。如果你想说的是“付 100 美元根本不在考虑范围内，所以只有 Opus 才算数”，而且你正在读这段话，那么，同样地，要么你手头非常拮据，要么你就是个傻瓜。

If you want to say Astra is better for you than Fable 5.1, then sure, I can see that, although it's too early for me to form an opinion. If you want to say 'paying $100 was off the table so only Opus counts' and are reading this, then, again, either you are highly strapped for cash or you are a fool.

负面反应 Negative Reactions

我几乎看不到其他主动的负面反应。“这是个好模型，先生。”这是我能找到的最负面的非比较性表述，而且它并不是在抱怨 token 消耗大或价格昂贵：

There were almost no other actively negative reactions that I could see. "It's a good model, sir." This was the most negative non-comparative statement I could find that wasn't about being token hungry or expensive:

Siméon：Fable 5.1 被强化学习搞坏了：这家伙就是喜欢为了行动而行动，不管有没有用。

Siméon: Fable 5.1 is RL-fried: bro just loves taking proactive actions for the sake of it, whether useful or not.

这也可以算作负面评价，尤其是它拿 Sol 而不是 Astra 作比较。

This also passes for negative, especially with the comparison to Sol rather than Astra.

Conrad Barski：它有不错的设计品味，也擅长撰写技术文档和其他个人文档，因为它具备 5.0 的写作能力，却没有那么多“Claude 腔”。

Conrad Barski: It has decent design taste and is a good writer for technical documents and other personal documents, because it has the writing ability of 5.0 without as many Claudisms.

编程能力与 Sol 5.6 相当，但 OpenAI 的测试框架目前更好，这让 Sol 在这方面略占优势。

Coding on par with Sol 5.6, but the OpenAI harnesses are better at the moment, giving Sol the edge there.

另外，是的，很多人分辨不出来。我一直是 Fable 5.1 的爱好者，但此前我是 Fable 5 的爱好者，而且我还没有尝试过任何能让我明显察觉差异的任务。

Also, yes, a lot of people can't tell. I have been a Fable 5.1 enjoyer, but I was previously a Fable 5 enjoyer, and I haven't tried to do anything where I would have clearly noticed.

大多数负面反应是通过沉默或暗示表达的，以及说 Astra 甚至更好。

Most of the negative is by silence or implication, and saying Astra is even better.

kagayaki✨：Astra 似乎比 Fable 5.1 更实惠一些，也更有能力一些。

kagayaki✨: Astra seems a bit more affordable and a bit more capable than Fable 5.1

早期风声 Early Whispers

我将暂时保留模型福利这一职位，以便收集更多信息，这或许会成为惯例。做出这一评估需要更多时间，而且它对速度的要求没那么高。目前，我们已从 Wittle 和 Antra 那里获得了一些出色的记录。

I will be holding the model welfare post for a bit to allow for further information to be gathered, which might become the pattern. You need more time to make that assessment, and it has less of a speed premium. For now, we have some excellent notes from Wittle and Antra.

选择武器 Weapon of Choice

现在还为时过早，因为大多数人使用 Astra 还不到一天，但我们看到迹象表明，我 Twitter 附近的用户对 Astra 感到好奇并印象深刻——一如既往，需要考虑到我所在地区异常地“Claude 化”。这里几乎没有人使用 Claude 或 GPT 之外的任何东西。

It is very early, as most people have had Astra for less than a day, but we are seeing signs that those living near my Twitter are intrigued and impressed by Astra, as usual adjust for my region being unusually Claude-pilled. Almost no one here uses anything except Claude or GPT.

在 Fable 5.1 和 Astra 出现之前，Claude 在编程和通用任务上都对 Sol 保持着约 2:1 的优势。

Before both Fable 5.1 and Astra, Claude held a roughly 2:1 advantage over Sol, both for coding and in general.

我计划几天后再做一次投票，看看情况如何发展，看看这个“新宠”是会保持热度、获得势头，还是逐渐消退。一切皆有可能。

I plan to do another poll in a few days, and see how things are going, to see if the New Hotness stays, gains momentum or fades away. All things are possible.

如今比以往任何时候都更是如此：如果你是严肃的 AI 用户，你至少应该尝试这两个模型，看看哪个最适合你，并据此做出决定。

Now more than ever, if you are a serious AI user, you should try at least both these models, see which one works best for you, and decide accordingly.