Summary
On July 24, 2026, Anthropic released Claude Opus 5, its new flagship model, positioning it as near Fable 5's frontier intelligence at half the price. Opus 5 pairs a 1M-token context window and up to 128k tokens of output with a new five-level effort control, and ships at the same $5/$25 per-million-token API pricing as Opus 4.8, rolling out on the Anthropic API and cloud platforms including Amazon Bedrock.
What changed
Anthropic released Claude Opus 5 on 2026-07-24 with a 1M-token context window, up to 128k synchronous output tokens, a new five-level reasoning effort control, and its strongest reported Opus benchmark profile. API pricing holds at $5 input / $25 output per million tokens (same as Opus 4.8), with a Fast mode at ~2.5x speed for $10/$50. Available via the Anthropic API and Amazon Bedrock at launch.
Why it matters
Holding flagship pricing flat while adding a granular effort dial lets teams trade cost against depth per request instead of switching model tiers, and pricing Opus 5 as 'frontier at half the price of Fable 5' directly pressures the top of the model market. The launch lands in the same late-July window as GPT-5.6, DeepSeek V4 GA, and Kimi K3, compressing the frontier price-performance gap.
Evidence excerpt
Claude Opus 5 combines a 1 million-token context window, up to 128,000 tokens of synchronous output, a new five-level effort control, and costs $5 per million input tokens and $25 per million output tokens on the Claude API, the same rate as Opus 4.8.