Claude Opus 4.8, released May 28, 2026, refines Opus 4.7 with stronger agentic judgment, improved coding honesty, and significantly cheaper Fast mode — while holding the same $5/$25 per-MTok price as its predecessors.
Why It's in Adopt
Opus 4.8 is the right Opus to run today for most production agentic workloads. The upgrade from 4.7 is incremental but meaningfully practical:
- Improved agentic judgment: In Claude Code, Opus 4.8 asks clarifying questions more effectively, catches its own mistakes before reporting completion, and pushes back when a plan is unsound. Less silent failure in complex multi-step tasks.
- 4× more honest about code flaws: Opus 4.8 is roughly four times less likely than 4.7 to let defects in code it has written pass unremarked — a direct quality-of-life improvement for code review and refactor workflows.
- Dynamic Workflows (research preview): A new feature designed to manage complex tasks across hundreds of parallel subagents. Still early, but the mechanism for long-running agent networks.
- Cheaper Fast mode: Fast mode drops from $30/$150 per MTok (Opus 4.6/4.7) to $10/$50 — a 3× reduction — making speed-sensitive agentic pipelines significantly more accessible.
Performance
| Benchmark | Opus 4.8 | Opus 4.7 |
|---|---|---|
| SWE-bench Verified | 88.6% | 87.6% |
| SWE-bench Pro | 69.2% | 64.3% |
| GPQA Diamond | 93.6% | 94.2% |
The +5 point gain on SWE-bench Pro (real GitHub issues) is the most relevant signal for engineering teams — it indicates stronger resolution of complex, multi-file tasks, not just incremental coding polish.
Fable 5, released June 9, 2026, scores higher still (SWE-bench Verified 95%) but carries premium pricing. Opus 4.8 remains the production baseline for teams who do not need the highest capability ceiling.
Pricing
| Property | Value |
|---|---|
| Input | $5 per million tokens |
| Output | $25 per million tokens |
| Fast mode input | $10 per million tokens |
| Fast mode output | $50 per million tokens |
| Context window | 1,000,000 tokens |
| Prompt caching | Up to 90% savings on repeated context |
| Batch processing | 50% savings for async tasks |
Fast mode pricing ($10/$50) is 3× cheaper than the same feature on Opus 4.6 and 4.7 ($30/$150).
How to Access
- Claude.ai — available on Pro and higher plans
- Anthropic API —
claude-opus-4-8model ID - Amazon Bedrock — available at GA
- Google Cloud Vertex AI — available at GA
- GitHub Copilot — available
Key Characteristics
| Property | Value |
|---|---|
| License | Proprietary |
| Context window | 1,000,000 tokens (1M) |
| Pricing | $5/M input, $25/M output |
| Provider | Anthropic |
| API model ID | claude-opus-4-8 |
| Release date | May 28, 2026 |