Anthropic releases Claude Fabel 5, it’s “Mythos-class” model.
There were rumors that Mythos would get a public release today and that's proven to be true. Anthropic is calling it Fable 5 and says it's a 'Mythos level' model.
It's been immediately rolled out and says it takes 2x the usage of Opus, so those rate limits will come fast.
Fable 5 is clearly positioned as the new top general-use model. It leads most benchmarks shown, especially agentic coding, knowledge work, spatial reasoning, tool use, legal, biology, cybersecurity, and health.
Here are the benchmarks it touts:
Reports say Claude Fable 5 and Claude Mythos 5 share the same underlying model, but Fable has stronger safeguards and this underscores it.
Here are the five killer applications, based on those benchmarks:
- Agentic coding: Best-in-table coding scores suggest it can run longer software tasks, debug complex codebases, and act more like an autonomous engineer.
- Knowledge work: Strong GDPval-AA performance points to better research, document synthesis, financial analysis, briefing notes, and complex professional reasoning.
- Computer use: Near-leading OSWorld results suggest it can operate apps, navigate workflows, fill forms, test software, and automate desktop tasks.
- Spatial reasoning: Big jump on Blueprint-Bench suggests stronger ability to interpret diagrams, plans, layouts, engineering drawings, and visual-spatial problems.
- Regulated professional domains: Strong legal, health, biology, and cybersecurity scores suggest useful expert-assistant applications, though likely with heavier safety constraints.
The initial previews of Mythos have been glowing, so this release is a big deal. I first wrote about Mythos shortly after it was leaked on March 30 and -- notably -- that was the the time when technology stocks bottomed. Since the, rumors of Mythos' power -- particularly in cybersecurity -- have been near constant.
At the time I wrote:
Finally, there has been no shortage of hype around a 'step change' in models and we've seen it so many times before. But if it's true and we are getting a new generation of truly superior models, that further extends the ceiling of what AI can do and how disruptive it is for the economy, and ultimately, how useful it will be.
Now we get the real test. I will dig into it and see what I can learn, at least on the financial side.
Up next, OpenAI was said to finish a training run on its latest model in March as well and it's code named Spud. It could be the next big iteration beyond this, or it could be playing catch-up.
This article was written by Adam Button at investinglive.com.提供 MainLink：Investinglive RSS Breaking News Feed
FX初心者には必須 無料のうちにGET！
