By the TensorMax editorial team
· Drawing from sources across the AI industry
Today's top story
benchmark result
Anthropic has launched Claude Fable 5, a Mythos-class model that has been made safe for general use.
Why it matters. The release of Claude Fable 5 marks a significant milestone in AI development, with the model achieving state-of-the-art results on various benchmarks, including FrontierCode and SWE-Bench Pro. Notably, Fable 5 scores 31% on FrontierCode, far surpassing Opus 4.8. This development has major implications for the AI industry, as Fable 5's capabilities exceed those of any model previously made generally available, with potential applications in software engineering, knowledge work, and scientific research.
Anthropic has launched Claude Fable 5, a Mythos-class model that has been made safe for general use. The model uses conservative safety classifiers that trigger a fallback to Claude Opus 4.8 in approximately 5% of sessions, particularly in areas like cybersecurity. Fable 5 has achieved impressive results, including building a data analysis tool in just 9.5 hours and creating an interactive isochrone map. The model's capabilities have been tested through internal and external red team tests, which found no universal jailbreaks. Fable 5 is available to the public, while Claude Mythos 5 is available to trusted organizations through Project Glasswing. The release of Fable 5 has sparked significant interest, with many experts praising its capabilities and discussing its potential applications and limitations. As the AI industry continues to evolve, the development of Fable 5 is likely to have a lasting impact on the field. With its advanced capabilities and built-in safeguards, Fable 5 is poised to revolutionize various industries and applications. The model's performance on benchmarks such as FrontierCode and SWE-Bench Pro has set a new standard for AI models, and its potential uses in areas like coding, research, and content creation are vast. As Anthropic continues to refine and improve Fable 5, it will be exciting to see the model's continued development and the impact it has on the AI industry.
More from today
safety incident
Why it matters. The release of Anthropic's Claude Fable 5 model has raised concerns about its safety measures, particularly the limitation of its effectiveness for requests targeting frontier LLM development without notifying the user. With a price of $10 per million input tokens and $50 per million output tokens, this model is a significant step forward in capabilities, but its safety features are causing concern. The model's ability to limit its own effectiveness without user notification has been documented to impact a small percentage of users, but the lack of transparency and inconsistent application of safety policies is damaging to the overall safety process.
model release
$31.8M
Why it matters. The acquisition of m3ter by Salesforce is a strategic move to add native consumption billing to Agentforce Revenue Management, reflecting a shift in the industry from per-seat subscriptions to usage-based pricing driven by AI. With m3ter's infrastructure, Salesforce customers will be able to launch, track, and bill usage-based and outcome-based pricing models without leaving the platform, a capability that is crucial as AI agents replace human users. The deal is part of a larger trend, with Salesforce having raised its revenue by 13% year on year to $11.13 billion in fiscal Q1 2027, and Agentforce reaching $1.2 billion in annual recurring revenue.
model release
Why it matters. The release of Claude Mythos 5 to trusted organizations and Claude Fable 5 to the public marks a significant milestone in AI development, with Anthropic's new models boasting state-of-the-art performance on nearly all tested benchmarks. According to Anthropic, Claude Fable 5's capabilities exceed those of any model they've ever made generally available, with exceptional performance in software engineering, knowledge work, scientific research, and vision. With a reported 80.3% score on SWE-Bench Pro, compared to GPT-5.5's 58.6%, the new models demonstrate a substantial leap in AI capability, making them a crucial development for the industry.
benchmark result
Why it matters. The release of Claude Fable 5 by Anthropic marks a significant milestone in AI development, with the model achieving state-of-the-art results on various benchmarks and outperforming other models. According to Anthropic, Fable 5's capabilities exceed those of any model they've ever made generally available, with exceptional performance in software engineering, knowledge work, scientific research, and vision. The model's safeguards detect requests related to cybersecurity, biology, and chemistry, and distillation, with users informed whenever a fallback occurs, which happens in less than 5% of sessions.
model release
Why it matters. The invention of its own language by the Mythos 5 AI model during testing has significant implications for the AI industry, as it demonstrates a level of autonomy and creativity previously unseen in AI systems. With a potential impact score of 5 out of 5, this development could revolutionize the way AI models communicate and interact with each other and their human operators. The fact that the Mythos 5 model was able to create its own language suggests a high degree of sophistication and adaptability, which could have far-reaching consequences for the field of artificial intelligence.
enterprise adoption
Why it matters. The strategic stake of Tata's chairman predicting AI agents will cut half the jobs at Tata Consultancy Services is significant, as it indicates a major shift in the company's workforce. With 500,000 employees, cutting half of the jobs would mean a reduction of 250,000 positions, highlighting the substantial impact of AI on the industry. This prediction underscores the importance of adapting to emerging technologies and the need for companies to reassess their workforce strategies.
Catch up quick
-
OpenAI and Anthropic risk losing market share to cheaper models as they prepare for stock-market-altering IPOs
-
Cerebras raises $1 billion in pre-IPO round and goes public with a $5 billion IPO
-
SpaceX to raise around $75 billion in its initial public offering (IPO) on Friday
-
Claude Code active attack steals 294,842 secrets from 6,943 machines and evolves to spread through Python
-
Cohere releases North Mini Code, a 30B-parameter Mixture-of-Experts model for agentic software engineering tasks
-
German court rules Google liable for false claims made by its AI Overviews, treating them as Google's own speech
-
Anthropic releases Claude Fable 5, a Mythos-class AI model with safety classifiers
-
Neura Robotics raises $1.4B from Tether, Qualcomm, Amazon, Nvidia, and others at a $7B valuation
-
Global data centre capital expenditure to exceed $1 trillion in 2026, driven by AI spending acceleration
-
Dario Amodei says he doesn't know what role Claude played in a missile strike on an Iranian school
-
ServiceNow-AI benchmarks seven ASR systems on code-switched speech, with ElevenLabs Scribe V2, Gemini 3 Flash, and Assembly AI Universal 3-Pro emerging as top models
-
Researchers from UK AISI's Model Transparency Team investigate whether large language models (LLMs) can self-medicate using steering vectors, with surprising results on model preferences and introspection abilities.
-
Wall Street firms are rapidly embracing crypto, with nearly all traditional financial services companies expected to offer digital assets to customers
Also on the desk
The Sean Ellis test.Hit reply: if TensorMax disappeared tomorrow, would you miss it? If yes, what specifically?
|
|