Anthropic has released Opus 5, its latest model update, which performs comparably to its more expensive Fable model on coding and development benchmarks while costing roughly half as much. The model achieves iterative performance gains rather than a breakthrough leap, and deliberately avoids cutting-edge cybersecurity training, leaving it substantially behind competitors like Mythos 5 on vulnerability exploitation.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Anthropic released Opus 5, an update to its Opus model, which recently became popular for coding and software development tasks. Benchmarks show Opus 5 performing at roughly the same level or slightly ahead of Anthropic's Fable model for coding tasks, and beating Opus 4.8 and OpenAI's GPT-5.6-Sol across most task types.
Why it matters
Opus 5's main value proposition is delivering performance close to Fable at approximately half the cost, making it an attractive option for developers seeking cost-efficient AI for development work. However, Anthropic deliberately limited Opus 5's training on cybersecurity tasks, leaving it substantially behind Mythos 5 on vulnerability exploitation, a trade-off that shapes its suitability for security-sensitive applications.
What to watch
Opus 5 lacks some of Fable's controversial protections, such as the 30-day data retention policy for incident review, signaling a different approach to safety trade-offs in the new model.
Anthropic released Opus 5 as the latest iteration of its Opus model, which has recently become a popular choice for coding and other software development tasks. The release represents an incremental update rather than a capability breakthrough.
Benchmark performance across multiple tests—including Frontier-Bench and DeepSWE—shows Opus 5 performing at about the same level or slightly ahead of Anthropic's Fable model on coding tasks. Across task types more broadly, Opus 5 beats Opus 4.8 and OpenAI's GPT-5.6-Sol. The improvements are consistent but incremental; the benchmarks reflect an iterative increase in performance rather than a radical leap.
Anthropoc's central pitch for Opus 5 is that it delivers performance close to Fable at approximately half the cost. This positions the model as a cost-efficient alternative for teams seeking strong development capabilities without the higher expense of the company's most advanced offering.
Notably, Anthropic made deliberate training decisions that shape Opus 5's profile. The company specifically avoided giving Opus 5 cutting-edge training on cybersecurity tasks, resulting in a significant performance gap. While Anthropic claims Opus 5 is relatively good at finding cybersecurity vulnerabilities, it is "substantially behind Mythos 5 on the exploitation of those vulnerabilities." This limitation reflects a trade-off: cost efficiency and ease of deployment over maximum security-related performance. Additionally, Opus 5 lacks some of Fable's safety protections, including Fable's 30-day data retention policy for incident review, indicating a different approach to oversight and risk management in the new model.
Anthropic's Opus 5 represents a deliberate pivot toward efficiency rather than raw capability. The benchmarks across Frontier-Bench, DeepSWE, and other metrics show steady iterative gains—Opus 5 outperforms Opus 4.8 and GPT-5.6-Sol across most task types while matching or slightly exceeding Fable on coding performance. However, the absence of a breakthrough in agentic coding performance (the kind of leap Opus 4.5 achieved) suggests the model is optimized for a different market positioning: developers who need strong performance on common tasks without the premium cost of Fable.
The cybersecurity limitation is noteworthy as a conscious engineering choice. By training Opus 5 without cutting-edge cybersecurity capabilities, Anthropic has intentionally ceded ground to models like Mythos 5 on vulnerability exploitation. This decision also correlates with a shift in safety guardrails—Opus 5 drops Fable's 30-day data retention policy and other protections, suggesting Anthropic views Opus 5 as a more cost-optimized product with different risk and oversight assumptions than its higher-end offering.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
