Claude Opus 4.6 introduced (1M context beta, 128k output, $5/$25 per MTok)
Anthropic introduced Claude Opus 4.6 with improved coding, long-horizon agentic work, and a 1M token context window (beta). Standard API pricing remains $5/$25 per million input/output tokens, with premium long-context pricing above 200K input tokens.
Impact summary
Use explicit levels instead of abstract charts to judge this change.
- New model id: claude-opus-4-6
- 1M token context window (beta) with premium pricing when input > 200K tokens
- Up to 128k output tokens
- New API controls: adaptive thinking + effort levels; context compaction (beta)
- Standard: $5 / MTok input, $25 / MTok output
- Long-context (>200K input tokens with 1M context enabled): $10 / MTok input, $37.50 / MTok output
- Batch: $2.50 / MTok input, $12.50 / MTok output
- Validate code generation/refactor quality and long-running agent behavior on your repo.
- If you plan to use 1M context, budget for premium long-context rates and verify your org eligibility (tier 4 / custom limits).
Recommended Actions
Only actionable items stay prominent. Incomplete actions fall back to a compact list.
Test Opus 4.6 on your main coding workflows (generation, refactor, review, debugging), and compare against your current baseline model.
If enabling 1M context, add cost guardrails for long-context premium pricing and validate org eligibility.
Sources
Each source focuses on what it confirms, without repeating the whole article.
Pricing remains the same at $5/$25 per million tokens; Opus 4.6 adds 1M context (beta) and 128k output tokens.
Long context pricing: Opus 4.6 standard $5/$25 per MTok; premium long-context above 200K input tokens $10/$37.50.