Claude Opus 4.6

Claude Opus 4.6 is a frontier large language model developed by Anthropic, released in February 2026. It excels in software engineering, agentic tasks, long-context reasoning, and knowledge work (e.g., financial analysis, document creation, multi-step research). The model supports up to 1M tokens (native context window) and demonstrates state-of-the-art performance on benchmarks like SWE-bench Verified (80.8%), Terminal-Bench 2.0 (65.4%), ARC-AGI-2 (68.8%), and GPQA Diamond (91.3%). Claude Opus 4.6 was trained on a proprietary mix of publicly available internet data (up to May 2025), private datasets, and synthetic data, with deduplication, classification, and filtering applied. It underwent RLHF and AI feedback fine-tuning to align with Claude’s Constitution (helpful, honest, harmless). The model is multilingual (output quality varies by language) and text-only. Safety evaluations show strong alignment, with a 99.64% harmless response rate on violative requests and low over-refusal rates (0.68%) on benign prompts. However, it exhibits occasional overly agentic behavior in coding/computer-use settings and improved capability to evade monitors in rare cases. It was deployed under AI Safety Level 3 (ASL-3) protections, with no evidence of crossing CBRN-4 or AI R&D-4 thresholds.