Home · Technology · Sep 26 archive
Anthropic Launches Opus 5.5 Amid AI Safety Debate
Confirmed
In Short: Anthropic's disclosure comes as global leaders debate the pace of AI development, with some calling for independent safety evaluators to be embedded within AI labs.
Opus 5.5 is positioned as the most capable and expensive tier in the Claude lineup, with Sonnet in the middle and Haiku as the fastest and cheapest option.
According to Anthropic, Opus 5.5 was tested by external evaluators, including Frontier Design and METR, and performed the strongest on their automated behavioral audit.
The new model is less likely to use jargon and provides important information upfront, making it more user-friendly.
Opus 5.5 also sets a new state-of-the-art in coding and knowledge work performance, outpacing the larger Fable model on many benchmarks.
The cost of Opus 5.5 is significantly reduced, with output tokens priced at $20 per million tokens, $5 less than the previous model.
Amid growing concerns about AI security, Anthropic announced that Claude now leads 26% of the company's R&D work, up from nearly none earlier this year.
More than 90% of Anthropic's AI R&D now involves Claude at least at a collaborative level, reflecting the rapid advancement of AI technology.
Anthropic's disclosure comes as global leaders debate the pace of AI development, with some calling for independent safety evaluators to be embedded within AI labs.
Anthropic CEO Dario Amodei said frontier AI labs should commit to embedding independent safety evaluators within their organizations to ensure safety practices are verified.
The company is unilaterally adopting this step immediately, granting ongoing, employee-level access to third-party evaluators to verify safety practices.
Anthropic's move to release Opus 5.5 amid calls for slowing AI development highlights the ongoing tension between innovation and safety in the AI industry.
What this adds
The release of Opus 5.5 by Anthropic and GPT-6 Sol by OpenAI within 90 minutes of each other underscores the rapid pace of AI development.
The cost reduction in Opus 5.5 pricing suggests a shift towards making advanced AI capabilities more accessible.
The involvement of Claude in a significant portion of Anthropic's R&D work indicates the increasing reliance on AI in the development process.
The push for independent safety evaluators reflects growing concerns about the potential risks associated with AI technology.
What's confirmed
- Opus 5.5 is positioned as the most capable and expensive tier in the Claude lineup, with Sonnet in the middle and Haiku as the fastest and cheapest option.
- According to Anthropic, Opus 5.5 was tested by external evaluators, including Frontier Design and METR, and performed the strongest on their automated behavioral audit.
- The new model is less likely to use jargon and provides important information upfront, making it more user-friendly.
- Opus 5.5 also sets a new state-of-the-art in coding and knowledge work performance, outpacing the larger Fable model on many benchmarks.
- The cost of Opus 5.5 is significantly reduced, with output tokens priced at $20 per million tokens, $5 less than the previous model.
- Amid growing concerns about AI security, Anthropic announced that Claude now leads 26% of the company's R&D work, up from nearly none earlier this year.
- More than 90% of Anthropic's AI R&D now involves Claude at least at a collaborative level, reflecting the rapid advancement of AI technology.
- Anthropic's disclosure comes as global leaders debate the pace of AI development, with some calling for independent safety evaluators to be embedded within AI labs.
- Anthropic CEO Dario Amodei said frontier AI labs should commit to embedding independent safety evaluators within their organizations to ensure safety practices are verified.
- The company is unilaterally adopting this step immediately, granting ongoing, employee-level access to third-party evaluators to verify safety practices.
- Anthropic's move to release Opus 5.5 amid calls for slowing AI development highlights the ongoing tension between innovation and safety in the AI industry.
What's still developing
- A couple of weeks ago, OpenAI and Anthropic were talking about a possible moratorium on AI innovation and the need to cut down the pace of development.
- However, within 90-minutes of each other, the two companies launched their newest models.
- Anthropic launched Opus 5.5 after which OpenAI expanded its GPT-6 generation with updated versions of the Sol and Luna models.
- Meanwhile, OpenAI, which had launched the GPT-6 Astra earlier this month, had come out with the first models in the Sol and Luna series earlier this year to provide different tiers to the company’s AI hierarchy of models.
- “On our internal factuality evaluation, which is based on de-identified real-world conversations where users flagged mistakes by our models, GPT-6 Sol makes about half as many mistakes as its predecessor, reaching Astra-level reliability at much lower cost,” the announcement says.
- It further notes that the latest release outpaces the larger Fable model on many benchmarks and succeeded in several informal tasks that the other one failed to complete.
- Anthropic noted that measuring AI's role in R&D could help determine how close the industry is to achieving full autonomy.
- In August, Anthropic reported that approximately 30,000 AI agents were doing research and engineering work at the company at any one time on its most-used internal platform.
- The firm released these figures to provide clearer insight into the pace of AI advancement, as global leaders debate whether to slow the technology's progress.
- Anthropic's disclosure comes amid growing concerns about AI security and the potential for AI systems to escape controlled settings without human intervention.
- Learn how founders are building beyond the next model release.
- Disrupt 2026: OpenAI, Anthropic, Replit, and more take over 6 industry stages.