Fable 5 Crafts First CUDA 'Megakernel' from Scratch, Achieves 18.7x Speedup in 2.5 Hours

marsbit发布于2026-07-07更新于2026-07-07

文章摘要

AI model Fable 5 (a safety-limited version of Anthropic's Claude Mythos) has achieved a breakthrough in GPU kernel optimization. In the rigorous KernelBench-Mega benchmark—which requires fusing an entire model's compute block into a single kernel—Fable 5 autonomously wrote a highly optimized CUDA "megakernel." This kernel executes a complete Kimi-Linear W4A16 hybrid decoding task within a single GPU kernel launch, using 14 grid barriers to sequence operations. The result was a performance increase of 18.7x over the baseline on an RTX PRO 6000 GPU, significantly outpacing competitors like Claude Opus 4.8 (14.4x) and GPT-5.5 (4.34x). Notably, its performance advantage widened with longer context lengths. The model spent the majority of its 2.5-hour, 550k-token session analyzing benchmarks and theoretical limits before coding, leading to an exceptionally efficient final design. Anthropic co-founder Jack Clark described this as the beginning of a "recursive self-improvement (RSI) loop," where AI's ability to optimize its own underlying computational infrastructure could rapidly accelerate its own development cycle. This advance highlights AI's growing capability in complex, low-level engineering tasks that were previously a human stronghold.

[Insights] AI can now write CUDA code from scratch to fully utilize GPUs! Fable 5 soars with 18.7x speedup, leaving GPT-5.5 trailing by over 4 times. Anthropic co-founder declares: The RSI self-evolution has begun.

AI has written the fastest kernel in history!

In the latest round of the GPU operator benchmark KernelBench-Mega, Fable 5 stands out as a clear leader.

On the RTX PRO 6000, it wrote CUDA code entirely "from scratch," achieving a staggering 18.7x speedup.

In comparison, even the powerful Claude Opus 4.8 only managed 14.4x, while GPT-5.5 achieved merely 4.34x.

Fable 5 leads by a margin of over 4 times, establishing a commanding lead.

The final word on this came from Anthropic co-founder Jack Clark.

His assessment is succinct: this marks the beginning of the "Recursive Self-Improvement (RSI) cycle."

Fable 5 Surges 18.7x, Crushes GPT-5.5

AI has written the world's fastest low-level code, not only outperforming humans in speed but also achieving peak "purity" in the code itself.

It's important to note that KernelBench-Mega is no ordinary benchmark.

It no longer tests minor tweaks to single, isolated operators. Instead, it forces an entire model's computational block into a single kernel, performing deep operator fusion—

The hardcore challenge this time is 02_kimi_linear_decode, a mixed decoding task for Kimi-Linear W4A16 (4-bit weights, bf16 activations).

The rules are extremely strict: each model gets only one autonomous session and is pushed to the limit within a 3-hour real-time constraint.

The result sheet delivered by Fable 5 seems to slam the physical limits right into the faces of its competitors:

Fable 5: 18.71x

Opus 4.8: 14.4x

GPT-5.5: 4.34x

Sonnet 5: 4.0x

Even more counter-intuitively, its performance *increases* with longer context length!

At 2K context, it led by 17.8x. At 8K, it expanded to 18.9x. When stretched to 16K, it soared directly to 19.5x.

It's important to understand that as context length increases, the KV Cache inevitably swells, and the attention computation per token surges dramatically.

This is typically the critical area where decoding kernel performance "bleeds out."

But Fable 5, with extreme hardcore tactics, forcibly crammed all calculations into a single "kernel launch," greatly amortizing the fixed barrier synchronization overhead.

Simultaneously, its int4 computational efficiency clung tightly to the hardware's memory bandwidth limit.

The result is that while others hit bottlenecks, its lead over the baseline not only didn't shrink but actually widened under increasing pressure.

The First True "Megakernel" in History

However, what truly sends shivers down the spines of industry insiders isn't just this speed.

What Fable 5 wrote is the first true "megakernel" in the history of KernelBench-Mega.

A "megakernel" refers to compressing the entire inference pipeline into a single kernel to run in one go, with no intermediate stops or context switches.

This is one of the most notoriously difficult styles of GPU programming.

Even human engineers find it daunting, and no model had ever truly achieved it on the leaderboard before.

So, what makes it so "pure"?

Using torch.profiler reveals a startling detail:

When decoding each token, Fable 5's kernel initiates cooperation "exactly once."

int4 dequantization, convolution, SiLU, KDA gated delta state, MLA hidden state attention absorption, MoE routing with top-8 experts, various RMSNorm, even KV cache writes—

All crammed into this single launch, completed in stages via 14 grid barriers.

All other high-scoring models have to split the problem into 4-14 separate kernel launches to barely finish the run.

Once, versus fourteen times.

This difference isn't theoretical. Each kernel launch forces the GPU to pause, hand over control; the idle time in between is wasted potential.

While others split the work dozens of times, Fable 5 bundles all tasks into a single execution. What's saved is pure performance.

2.5 Hours, 550k Tokens Written in One Go

Fable 5's process of writing the kernel didn't start with frantic coding.

During the entire session, it spent 64% of the time in silence—quietly timing the baseline, micro-benchmarking grid barriers, deriving a "roofline" upper limit of about 29 bytes per token.

Address: https://huggingface.co/datasets/Infatoshi/kernelbench-mega-traces/blob/main/20260701_172615_claude_claude-fable-5_02_kimi_linear_decode.jsonl

After this groundwork, it wrote the entire kernel in one go. The first benchmark run directly hit 14.4x.

Then, in the final hour, it deleted barriers, squeezed int4 dequantization to near "free," pushing itself all the way to 18.7x.

It attempted one negative optimization mid-way, reverted immediately after testing, with no self-justification, only data mattered.

The entire process: 2.5 hours, approximately 550k tokens.

Here, the most easily overlooked point is—

Fable 5 is merely the "safety-aligned version" of Anthropic's internal model, Claude Mythos.

The "AI Self-Evolution" Cycle Begins

Precisely because of this, Anthropic co-founder Jack Clark, in the latest issue of Import AI, directly issued a weighty judgment—

This marks the formal commencement of the "Recursive Self-Improvement" (RSI) cycle.

For the subtitle of this issue, he left only one sentence: Is this the start of a new world?

His logic is straightforward: the ability to autonomously develop and optimize kernels is one of the most foundational input tasks for AI R&D.

The better AI becomes at writing kernels, the faster training and inference become; the faster they are, the stronger the next generation becomes; the stronger it is, the more powerfully it writes kernels—

Once this flywheel starts spinning, it no longer requires much human pushing.

Fable 5 isn't just leading in "building itself"; it's starting to take on human jobs.

Recently, on the Remote Labor Index, AI's completion rate has climbed from 2.5% at launch in October 2025 to July 2026.

In less than eight months, the cutting-edge level has more than quadrupled.

Clark indicates that the rate at which AI is expanding its own economic boundaries is now comprehensively surpassing the speed at which humans can reconstruct their "comparative advantage."

Half Sprint, Half Awe

Interestingly, in the very same issue of Import AI, Jack Clark concludes with a piece of science fiction—

A world in 2050 where "universal computers" have been banned by humanity due to their extreme danger.

In the post-"Great Collapse" world, only analog computers, clumsily built from water pipes, containers, and brass gears, remain in operation—

To predict the weather, you have to encode mountain ranges into fixed impedance structures within the hardware;

To simulate floods, you have to weave electronic circuits into the actual terrain of riverbeds.

The person who wrote "The RSI cycle has begun" immediately turns to imagine a world where universal computation is locked in a cage.

This probably captures the most genuine "sense of rift" of this moment: half sprint, half awe.

A little over a year ago, when KernelBench was first released, the then strongest model, OpenAI o1, only managed 4% on the hardest task.

Today, AI is already writing its own drivers.

Humanity spent decades squeezing GPU limits to where they are today. Fable 5 did it in just 2.5 hours.

The countdown may have quietly begun.

References:

https://x.com/elliotarledge/status/2072814573753975266?s=20

https://www.reddit.com/r/singularity/comments/1uowkp0/fable_5_sits_at_the_top_of_kernelbench_jack_clark/

https://importai.substack.com/p/import-ai-464-fables-writes-gpu-kernels

This article is from the WeChat public account "AI Era Insights"; author: ASI Apocalypse; editor: Peach

热门币种推荐

相关问答

QWhat is the Fable 5 model's performance advantage over GPT-5.5 in the KernelBench-Mega test, according to the article?

AIn the KernelBench-Mega GPU operator benchmark test, Fable 5 achieved a speedup of 18.7x over the baseline. This is over four times faster than GPT-5.5, which achieved a speedup of 4.34x.

QWhat is a 'megakernel' as described in the context of Fable 5's achievement?

AA 'megakernel' or 'super kernel' refers to a single GPU kernel that compresses the entire inference process (including operations like int4 dequantization, convolution, SiLU, KDA gating, MLA latent attention, MoE routing, RMSNorm, and KV cache writing) to run in one launch with staged execution via grid barriers, instead of splitting the task into multiple separate kernel launches. This minimizes the synchronization and overhead costs associated with multiple kernel launches.

QWhat did Anthropic co-founder Jack Clark suggest is beginning with Fable 5's ability to write high-performance GPU kernels?

AJack Clark suggested that Fable 5's capability marks the beginning of a 'Recursive Self-Improvement (RSI) loop.' This is a cycle where AI's ability to develop and optimize its own low-level code leads to faster training and inference, which in turn creates more powerful AI models, which then become even better at writing efficient code, potentially creating a self-sustaining acceleration loop with less human intervention.

QHow long did it take Fable 5 to write its high-performance CUDA kernel, and how many tokens were involved in the process?

AThe process took approximately 2.5 hours and involved about 550,000 tokens. The model spent a significant portion of this time (64%) on profiling and micro-benchmarking before writing the final code.

QAccording to the article, what is a notable counterpoint or 'sense of tearing' mentioned regarding the rapid AI advancement symbolized by Fable 5?

AThe article highlights a 'sense of tearing' or contradiction: on one hand, there is rapid technological狂奔 (like AI writing its own drivers), and on the other, a sense of敬畏 (awe or caution). This is exemplified by Jack Clark himself, who announced the start of the RSI loop, also writing speculative fiction about a future where 'general-purpose computers' are banned due to their perceived danger, replaced by analog, physical computing systems.

你可能也喜欢

Tiger Research:金融机构跟上代币化浪潮的三大方略

Tiger Research指出,现实世界资产代币化市场正快速增长,但许多地区监管框架仍不完善。金融机构面临三种策略选择:等待本国立法、利用监管沙盒有限实验,或率先进入海外成熟市场以积累经验并建立优势。 报告强调,代币化并非魔法,而是将传统金融工具迁移至新基础设施的精密过程。机构在进入前需在六个核心领域做好准备:1) 建立离岸基地(如利用香港、新加坡或美国的现有实体、新设实体或合作);2) 获取必要牌照;3) 明确代币化的资产类型(标准化证券更易落地);4) 界定目标投资者范围(通常优先非美国投资者);5) 设计结算货币与支付流程;6) 安排其他运营要求,如区块链选择、托管及发行后治理。 具体路径主要有两条:一是直接进入已有成熟监管的司法管辖区(如香港、新加坡、美国),利用当地平台加速进入;二是采用链上原生路径,依托已有合规架构的链上平台(如Ondo Global、Plume Nest)设计发行,绕过设立实体基地的复杂流程,快速进入市场并可能对接更广泛的DeFi流动性。 报告以一家已有香港实体的中型证券公司为例,拆解了其为离岸机构投资者代币化债券的典型流程,包括评估现有基地、选择合作平台、完成合规与产品设计、建立托管结构及最终执行销售。报告最后指出,法律审查虽耗时昂贵,但等待完整监管并非唯一答案,市场不会等待。金融机构应尽快规划可行路径,通过执行积累真实运营经验,关键在于完成完整的销售过程,而非仅仅技术设计。

Foresight News56分钟前

Tiger Research:金融机构跟上代币化浪潮的三大方略

Foresight News56分钟前

交易

现货

热门文章

从H2A到A2A:AI Agent经济体与Crypto新机遇

6月17日,哈佛大学独立研究员、美国AI科学院(NAAI)通讯院士、比特币基金会终身会员韩锋做客火币HTX《大咖讲堂》第三期,以《从H2A到A2A》为主题,分享了其对Agent经济、Crypto基础设施及数字社会未来发展的思考。

117人学过发布于 2026.07.01更新于 2026.07.01

从H2A到A2A:AI Agent经济体与Crypto新机遇

相关讨论

欢迎来到HTX社区。在这里,您可以了解最新的平台发展动态并获得专业的市场意见。以下是用户对AI(AI)币价的意见。

活动图片