Claude Opus 4.8
Anthropic's current Opus-tier flagship, released 27 May 2026 — same rate card as 4.7, better benchmarks, and a fast mode that got three times cheaper.
Claude Opus 4.8 specs at a glance
| Maker | Anthropic |
| Released | 27 May 2026 |
| Context window | 1,000,000 tokens |
| Max output | Up to 128,000 tokens |
| API pricing | $5 / $25 per 1M input / output tokens |
| Fast mode | $10 / $50 per 1M tokens · up to 2.5x output speed |
| Reasoning | Adaptive thinking · user-controllable effort levels |
| Best at | Long-horizon agentic work, coding, knowledge work |
Calculate API costs for this model → Compare its monthly bill against 15 other rate cards.
What's new in Claude Opus 4.8
Opus 4.8 landed just six weeks after Opus 4.7 and builds on it with improvements across benchmarks. Anthropic positions it as a more effective collaborator — clearer writing, stronger long-horizon agentic execution, and better code review and debugging.
Two additions stand out. First, users on claude.ai now get direct control over the amount of effort Claude puts into a task — a knob previously only exposed to API developers. Second, Claude Code gained a "dynamic workflows" feature: Claude can plan a very large task and then run hundreds of parallel subagents in a single session, which changes what an overnight coding run can realistically cover.
Fast mode: 2.5x speed, now 3x cheaper
Fast mode runs the same Opus 4.8 model at up to 2.5x the output tokens per second, at premium pricing of $10 / $50 per 1M tokens. That is three times cheaper than fast mode cost on previous Opus models, which makes latency-sensitive production use — interactive coding, live assistants — a much easier sell.
Pricing
The standard rate card is $5 per 1M input tokens and $25 per 1M output tokens — unchanged since Opus 4.5, four releases ago. Opus 4.8 uses the same tokenizer introduced with Opus 4.7, so if you already migrated to 4.7 there is no token-count surprise; if you are coming from 4.6 or earlier, expect the same input text to tokenize into more tokens.
Who should use Claude Opus 4.8
- Teams already on Opus 4.7 — same price, same API surface, better results. This is a straightforward upgrade.
- Long-running agentic workloads — the dynamic-workflows / parallel-subagent capability is aimed squarely at large refactors and overnight runs.
- Latency-sensitive products — the repriced fast mode makes 2.5x speed affordable.
- Anyone who does not need Fable-tier capability — Claude Fable 5 is stronger but costs 2x as much.
How Claude Opus 4.8 compares
- Claude Opus 4.8 vs Opus 4.7 — should you upgrade? →
- Claude Fable 5 vs Opus 4.8 — is the 2x price worth it? →
- Claude Sonnet 5 overview →
- GPT-5.5 ("Spud") overview →
Frequently asked questions
When was Claude Opus 4.8 released?
Anthropic released Claude Opus 4.8 on 27 May 2026, six weeks after Opus 4.7. It is the current Opus-tier flagship.
How much does Claude Opus 4.8 cost?
Claude Opus 4.8 costs $5 per million input tokens and $25 per million output tokens — the same rate card that has held since Opus 4.5. Fast mode costs $10/$50 per million tokens.
What is Claude Opus 4.8's context window?
Claude Opus 4.8 has a 1 million token context window and can produce up to 128,000 output tokens — the same as Opus 4.7.
What is fast mode?
Fast mode runs the same Opus 4.8 model at up to 2.5x the output speed for $10 per million input tokens and $50 per million output tokens. That is three times cheaper than fast mode was on previous models.
What's new vs Opus 4.7?
Improvements across benchmarks, user-controllable effort on claude.ai, and a 'dynamic workflows' feature in Claude Code that can run hundreds of parallel subagents in a single session.
Is Opus 4.8 Anthropic's most capable model?
It is the most capable Opus-tier model, but Claude Fable 5 — released June 2026 at $10/$50 — sits above it in raw capability.