Anthropic Claude Opus 4.8 official launch
Key Questions
What new features are included in Claude Opus 4.8?
Claude Opus 4.8 introduces Dynamic Workflows for parallel subagents and Effort Control for adjusting thinking depth. It also shows 4x improvement in catching its own code flaws along with enhanced honesty alignment.
How does Opus 4.8 perform compared to GPT-5.5 and Gemini 3.1 Pro?
Benchmarks indicate Opus 4.8 surpasses both GPT-5.5 and Gemini 3.1 Pro. It achieves 69.2% on agentic coding versus GPT-5.5's 58.6% and scores 84% on Mind2Web.
What is the context window size for Claude Opus 4.8?
Anthropic launched Claude Opus 4.8 with a 2M token context window.
Has pricing changed for Claude Opus 4.8?
Pricing remains unchanged from previous versions, though Fast Mode is now 3x cheaper.
What improvements does Opus 4.8 offer for coding and agents?
The model delivers stronger agentic coding performance and better self-detection of code flaws. It also supports enhanced agent skills and reasoning capabilities.
When will Mythos-class models become available to all customers?
Mythos remains restricted for now but Anthropic has teased broader access for all customers in the coming weeks.
How fast was the release cycle for Opus 4.8?
Opus 4.8 arrived just 41 days after version 4.7, reflecting Anthropic's rapid release cadence.
What makes Claude Opus 4.8 more honest than prior models?
Anthropic emphasizes honesty alignment in Opus 4.8, positioning it as the most honest model yet with reduced tendencies to pretend or withhold information.
Anthropic launched Opus 4.8 with Dynamic Workflows for parallel subagents, Effort Control for thinking depth, 4x better at catching its own code flaws, and honesty alignment. Benchmarks claim surpasses GPT-5.5 and Gemini 3.1 Pro; agentic coding 69.2% vs GPT-5.5 58.6%, Mind2Web 84%. Pricing unchanged, Fast Mode cheaper (3x). Mythos still restricted but teased for all customers in coming weeks. Fast release cycle (41 days after 4.7).