Anthropic just shipped Claude Opus 4.8 on May 28, 2026 — and it landed at #1 on the Artificial Analysis Intelligence Index with a score of 61.4, beating GPT-5.5 and Gemini 3.1 Pro.
Six weeks after Opus 4.7. That pace of release tells you everything about how competitive the AI race has gotten.
Here's what actually changed, what the benchmarks mean, and whether you need to upgrade.
What's New in Claude Opus 4.8
1. Massive Coding Improvement
The headline number: SWE-bench Pro score jumped from 64.3% to 69.2%.
That's a nearly 5-point gain in software engineering capability in six weeks. For context, GPT-5.5 sits at around 63% on the same benchmark.
Anthropic also says Opus 4.8 is roughly 4x less likely than 4.7 to let flaws in its own code pass without flagging them — a critical improvement for production use.
2. Math Performance Leap
96.7% on USAMO 2026, up from 69.3% on Opus 4.7.
That's not a typo. A nearly 28-point jump on competition-level mathematics. This puts Claude Opus 4.8 in a completely different tier for quantitative reasoning.
3. Fast Mode (2.5× Speed)
Opus 4.8 ships with Fast Mode as a research preview on the API. It delivers up to 2.5× more output tokens per second from the same model.
The pricing is now 3× cheaper than Fast Mode was for previous models — making speed-critical applications newly viable on Opus.
4. Dynamic Workflows in Claude Code
Claude Code gets a new Dynamic Workflows feature that lets it tackle very large-scale engineering problems — breaking them into subtasks, executing them in sequence, and managing state across the full problem.
This is Anthropic's answer to agents: not a separate product, but built into the core model.
5. Extended System Prompt Support
Opus 4.8 now accepts role: "system" messages after a user turn in the messages array. This means you can append updated instructions mid-conversation without restating the full system prompt — a significant quality-of-life improvement for long-running agentic applications.
Benchmark Comparison
| Benchmark | Opus 4.7 | Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro | |-----------|----------|----------|---------|----------------| | SWE-bench Pro | 64.3% | 69.2% | 63.1% | 61.8% | | USAMO 2026 (Math) | 69.3% | 96.7% | 88.2% | 84.1% | | AI Index Score | 58.1 | 61.4 | 59.7 | 58.9 |
Claude Opus 4.8 takes the top spot across all three key benchmarks.
Pricing
Pricing is unchanged from Opus 4.7:
- Input: $5 per million tokens
- Output: $25 per million tokens
- Fast Mode: Available via API, 3× cheaper than previous fast modes
For consumer access, Claude Pro ($20/mo) gives you Opus 4.8 alongside Claude Sonnet and Fable.
Should You Upgrade?
If you're using Claude Opus 4.7: Yes, immediately. Same price, meaningfully better coding and math. No reason to stay on 4.7.
If you're on Claude Sonnet: Depends on your use case. Sonnet is faster and cheaper. For coding and complex reasoning, Opus 4.8's gains are substantial.
If you're on GPT-5.5: The benchmark gap is now real. Worth testing Claude Opus 4.8 on your specific workloads — especially if you do a lot of code or math.
Verdict
Claude Opus 4.8 is the best AI model available right now by the numbers. The coding and math improvements are exceptional, Fast Mode makes speed-sensitive use cases viable, and Dynamic Workflows unlock a new category of large-scale agentic tasks.
Anthropic is shipping fast. GPT-5.6 and Gemini 3.5 Pro are both expected before July — the battle for #1 will be decided in the next few weeks.
Rating: 9.5/10
Try Claude Opus 4.8 on Claude.ai → Access via Anthropic API →
Published June 18, 2026. Based on Anthropic's official release notes and third-party benchmark data.
Sources: