Anthropic releases lower-cost Opus 5 AI model
Developed over time first seen 2 months ago
Anthropic has released Opus 5, a new heavyweight AI model it says is cheaper, less restrictive and stronger than its larger Fable 5 model on several benchmarks. The company says it is better at verifying and refining its own work, positioning it as a potentially more practical choice for many users while reducing interruptions from safety controls.
The launch follows Opus 4.8 in May and the June releases of Mythos 5, Fable 5 and Sonnet 5; Haiku remains the only model awaiting a fifth-generation update. Opus 5 is not subject to the 30-day data-retention policy affecting Fable and Mythos, though cybersecurity protections remain; Anthropic expects safety classifiers to activate 85% less often than for Fable 5 and is testing automatic fallback routing when prompts are blocked.
- Anthropic launches cheaper, less restrictive Opus 5.
- Opus 5 outperforms Fable 5 on some benchmarks.
- New fallback tool aims to reduce blocked requests.
New here? Start with this
Anthropic is an artificial intelligence company that makes Claude, a system designed to generate text, analyse information and help with tasks such as coding. Its models are offered in several versions with different strengths, prices and intended uses; Opus is generally its highest-capability range, while Sonnet and Haiku are designed for faster or lower-cost work.
AI companies regularly release new model generations and compare them using tests known as benchmarks, which measure performance on defined tasks. Such comparisons can indicate where a model may be useful, but results can vary depending on the task and how the system is used.
The company also places limits on some requests to reduce risks such as cyber abuse, and handles user data under differing policies for different products. These choices matter to businesses and developers because they affect cost, privacy, reliability and which tasks an AI system can be used for.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters argue that a cheaper, less restrictive model with stronger self-checking could make advanced AI more useful to a wider range of people and organisations. They may see reduced unnecessary safety-classifier triggers, no routine 30-day retention policy and fallback routing as practical improvements that preserve cybersecurity protections while limiting disruption to legitimate work.
The case against
Critics argue that rapidly releasing increasingly capable and less restrictive systems can weaken important safeguards before their real-world effects are understood. They may contend that lower intervention rates and reduced data retention make abuse, failures and accountability harder to detect, especially when model releases are arriving in quick succession.
More coverage
- Engadget — Anthropic says Opus 5 can nearly match its top-performing model for half the price
- The Verge — Anthropic releases Opus 5 with ‘close’ to Fable 5’s capabilities
Read the full article at the source →
Originally published by TechCrunch as “Anthropic launches Opus 5”.