Claude Opus 5 Brings Near-Fable 5 Coding to Half the Price

TL;DR

Anthropic dropped Claude Opus 5 on July 24 — a model that scores within 0.5% of Fable 5 on coding benchmarks while costing the same as Opus 4.8 ($5/$25 per million tokens). It is already live in GitHub Copilot, Cursor, and Claude Code. If you have been holding back on Fable 5 because of the $10/$50 pricing, this is the model you have been waiting for.

What Changed

Near-Fable 5 performance at Opus prices. On CursorBench, Opus 5 scores within 0.5% of Fable 5's peak — at half the cost per task. On Frontier-Bench, it more than doubles Opus 4.8's score. And on AA-Briefcase (agentic knowledge work tasks), it actually beats Fable 5 by approximately 150 Elo while cutting cost per task by 20%.

It is already in your tools. GitHub Copilot added Opus 5 support on launch day, describing it as suited for "complex, long-running coding tasks that require careful reasoning, effective tool use, and reliable execution across multiple steps." Cursor, Claude Code, and the Anthropic API all have it available now.

The effort parameter is the real unlock. Opus 5 introduces adjustable effort levels (low, medium, high, xhigh). Crank it to xhigh for hard refactoring and multi-file agent work. Drop it to low for autocomplete and linting. You control the cost-quality tradeoff directly.

Why This Matters

For the past month, developers faced a frustrating choice: pay Fable 5's $10/$50 per million token pricing for frontier coding intelligence, or settle for Sonnet 5's lower quality. Opus 5 closes the gap — you get frontier-adjacent coding at mid-tier prices.

The bigger story is Anthropic's release velocity. They shipped four Claude models in under two months: Opus 4.8, Sonnet 5, Fable 5, and now Opus 5. The release cadence is accelerating, and the performance gap between "flagship" and "workhorse" tiers is collapsing. Developers who build workflows around a single model should plan for frequent rotation.

This also resets the competitive landscape. Fable 5 was the coding benchmark leader, but its pricing limited adoption. Opus 5 brings that level of capability to a much wider audience — which puts pressure on OpenAI (GPT-5.6 Sol) and Google (Gemini 3.5 Flash Cyber, released three days earlier) to show their mid-tier models can keep up.

What To Do Now

1. Switch your Claude Code model to Opus 5. If you are on Sonnet 5 or Opus 4.8, the upgrade in coding quality is substantial with zero cost increase. Run claude config set model claude-opus-5 in your terminal.
2. Test xhigh effort for complex tasks. The default "high" is good for daily work, but xhigh is where Opus 5 shows its real capability on multi-file refactors and agentic debugging sessions.
3. Use the batch API for CI/CD pipelines. At $2.50/$12.50 per million tokens, Opus 5 batch pricing makes it viable for automated code review, test generation, and large-scale refactoring jobs.
4. Re-evaluate your Fable 5 dependency. If you are using Fable 5 primarily for coding, Opus 5 gives you approximately 99.5% of the performance at 50% of the cost. The only reason to stick with Fable 5 is for tasks that genuinely demand the absolute frontier — and most day-to-day coding does not.