Anthropic's Claude Opus 4.8 boosts coding benchmarks and dynamic workflows
New model promises honesty, scalable agent swarms, and 2.5x faster responses.
Deep Dive
Anthropic released Claude Opus 4.8, their strongest coding model yet, improving SWE-bench Pro from 64.3% to 69.2%. Priced unchanged at $5/$25 per million tokens, it introduces an effort parameter, a fast mode (2.5x speed at $10/$50) in research preview, and dynamic workflows in Claude Code that can fan out tasks to tens or hundreds of parallel subagents. The model emphasizes honesty and reduced misal
Key Points
- SWE-bench Pro improved from 64.3% to 69.2%, with explicit honesty about uncertainty and self-debugging.
- Fast mode offers 2.5x throughput at $10/$50 per million tokens (vs. $30/$150 for Opus 4.7 fast mode).
- Dynamic workflows in Claude Code enable hundreds of parallel subagents to collaboratively solve large tasks.
Why It Matters
Opus 4.8 brings scalable agent orchestration and honest AI reasoning, critical for trust in enterprise automation.