Anthropic launches Claude 4.7 Opus — the 1M-context tier becomes the standard agentic backbone, pricing held flat
Anthropic launched Claude 4.7 Opus this week, with the 1M-token context window now standard across all API tiers and pricing held flat against Claude 4.6 Opus. The flat pricing is the strategic move. Anthropic is betting that incremental capability gains delivered at constant per-token cost beat capability gains amortized via a price hike — a different bet than OpenAI's GPT-5 launch in March.
Claude 4.7's headline gains are agentic-task throughput (38% faster wall-clock on long multi-tool runs) and reduced hallucination on retrieval-heavy tasks (SimpleQA improvement from 0.71 to 0.79). The wall-clock improvement matters because agent runtime cost scales with wall-clock for most enterprise billing structures.
Holding pricing flat at the 1M-context tier is the structural play. Anthropic is making 1M-context the default expectation for the agent runtime category. Competitors that charge a premium for context-extended tiers (OpenAI, Mistral) now face the same context-window-as-floor pricing pressure that Claude has been pushing on the cached-context tier since Q1.
Anthropic — Claude 4.7 Opus release notes → · The Information — Claude 4.7 holds pricing flat → · TechCrunch — Claude 4.7 Opus launches with 1M-context default →