Anthropic's Claude Opus 4.8 boosts coding, cuts costs by 66%, matches Mythos alignment

Anthropic's latest flagship model, Claude Opus 4.8, launched on May 28, 2026, replacing Opus 4.7 at the same price point. The company claims the new model offers faster thinking modes for one-third the cost of its predecessor, while improving coding abilities—though it still trails OpenAI's GPT-5.5 on coding benchmarks. Notably, Anthropic emphasizes that Opus 4.8 reaches 'new highs on measures of prosocial traits like supporting user autonomy and acting in the user's best interest,' with misalignment rates substantially lower than Opus 4.7 and comparable to the company's Mythos Preview model. Opus 4.7 had a 92% honesty rate, and Anthropic appears to be raising the bar on safety and interpretability.

Alongside Opus 4.8, two other major releases from May 2026 stand out. OpenAI's GPT-5.5 Instant, launched May 5, replaces GPT-5.3 Instant as the default model in ChatGPT, boasting 52.5% fewer hallucinated claims on high-stakes prompts in medicine, law, and finance. Nvidia's Nemotron 3 Nano Omni, released April 28, is an open model that provides agents with multimodal input—unifying vision, audio, and text in a single perception-to-action loop, eliminating the need for separate models. These releases highlight a broader industry trend: balancing performance, cost efficiency, and safety while pushing toward more integrated agentic systems.