Claude Opus 4.8 Beats GPT-5.5 on Coding — And Runs 3x Cheaper
Anthropic released Claude Opus 4.8 on May 28, 2026 — its most capable model yet, scoring 69.2% on SWE-Bench Pro to outperform both GPT-5.5 and Gemini 3.1 Pro on the hardest software-engineering benchmark available. The release arrived simultaneously with Anthropic’s $65B Series H funding round, signaling the company is moving faster than ever. Crucially, a new “fast mode” runs at 2.5x the speed of Opus 4.7 at one-third the cost — making frontier-level AI significantly more accessible to developers who need both quality and throughput.
What Is Claude Opus 4.8?
Claude Opus 4.8 is Anthropic’s flagship large language model, the fourth major update in the Opus 4 series. It is designed for demanding agentic tasks — long-horizon coding sessions, complex reasoning chains, and autonomous multi-step workflows. Unlike previous Opus releases that focused primarily on raw benchmark performance, Opus 4.8 introduces architectural improvements to honesty and calibration: it is around four times less likely than its predecessor to let flaws in code pass unacknowledged. Anthropic describes it as their “most honest model yet.”
Key Features of Claude Opus 4.8
Fast Mode (3x cheaper): Opus 4.8 introduces a dedicated fast mode that delivers 2.5x the throughput of Opus 4.7. Pricing for fast mode is $10 per million input tokens and $50 per million output tokens — exactly one-third what Opus 4.7 cost at full speed. Standard mode pricing remains unchanged at $5/$25 per million tokens. For teams running large agentic pipelines, this is a significant cost reduction.
Dynamic Workflows in Claude Code (Research Preview): Perhaps the most consequential new capability. Claude Code can now plan a task, spawn hundreds of parallel subagents, and coordinate them within a single session. This effectively turns Claude Code into a distributed autonomous engineering team rather than a single assistant. Early users report being able to complete week-long refactors in hours.
Effort Control on Claude.ai: Users on Claude.ai and Cowork can now dial in how much compute Claude applies to a response. Lower effort = faster, cheaper responses. Higher effort = deep, deliberate reasoning. This allows the same model to serve quick Q&A and complex research tasks without switching interfaces.
Improved Honesty and Code Auditing: Opus 4.8 is 4x less likely to approve buggy code silently. Anthropic trained the model explicitly to surface its own errors, making it far safer to use in automated pipelines where human review is minimal.
How Claude Opus 4.8 Compares to Rivals
| Benchmark | Claude Opus 4.8 | GPT-5.5 | Gemini 3.1 Pro |
|---|---|---|---|
| SWE-Bench Pro | 69.2% | ~66%* | ~64%* |
| SWE-Bench Verified | 88.6% | ~87%* | ~85%* |
| Terminal-Bench 2.1 | 74.6% | ~76%* | ~72%* |
*Third-party estimates. GPT-5.5 leads on Terminal-Bench 2.1; Claude leads on SWE-Bench Pro.
Claude Opus 4.8 is also available natively in GitHub Copilot as of May 28, 2026, marking the first time an Anthropic model has shipped day-one in Microsoft’s developer ecosystem.
Access and Pricing
Claude Opus 4.8 is available now through:
- Claude.ai Pro and Team plans — included at no additional cost
- Anthropic API — $5/$25 per million tokens (standard); $10/$50 (fast mode)
- GitHub Copilot — available to all Copilot subscribers (Individual, Business, Enterprise)
- Amazon Bedrock and Google Cloud Vertex AI — model ID:
claude-opus-4-8
The Anthropic API also provides access to the Dynamic Workflows research preview for qualifying enterprise accounts.
What This Means For You
For individual developers: Claude Opus 4.8’s fast mode changes the cost calculus of using frontier AI in production. At 3x cheaper, building applications on top of Opus is no longer a budget concern for most teams. The honesty improvements also mean less manual auditing — a real productivity gain for solo builders.
For enterprises: Dynamic Workflows in Claude Code are the standout feature. The ability to delegate a large engineering task to hundreds of parallel subagents within a single Claude Code session is genuinely new territory. Anthropic is positioning this as the first step toward fully autonomous software development — a capability no competitor has publicly shipped at this scale.
For the AI industry broadly: Claude Opus 4.8 arrived on the same day Anthropic announced $65B in new funding at a $965B valuation — surpassing OpenAI’s most recent valuation. The AI race has two clear frontrunners, and the gap between them is narrowing to benchmark-by-benchmark marginal differences. Which model you choose increasingly comes down to workflow, not quality.
Frequently Asked Questions
Is Claude Opus 4.8 better than GPT-5.5?
On SWE-Bench Pro (the hardest coding benchmark available), Claude Opus 4.8 scores 69.2% vs GPT-5.5’s estimated ~66%, making Opus 4.8 the stronger choice for agentic coding tasks. GPT-5.5 leads on Terminal-Bench 2.1. For most development use cases, the two models are extremely close, and the choice often comes down to ecosystem (OpenAI/Microsoft vs Anthropic/GitHub).
How much does Claude Opus 4.8 cost?
Standard mode: $5 per million input tokens, $25 per million output tokens — same as Opus 4.7. Fast mode: $10/$50 per million tokens, but runs at 2.5x the speed and is 3x cheaper per unit of work completed relative to Opus 4.7’s throughput cost.
What is Dynamic Workflows in Claude Code?
Dynamic Workflows is a research-preview feature in Claude Code that allows Claude to autonomously plan a complex task, create hundreds of parallel subagents, and coordinate them all within one session. It’s designed for large-scale engineering tasks — refactoring codebases, running parallel test suites, or coordinating multi-file changes simultaneously.
What is Effort Control on Claude.ai?
Effort Control is a new slider in Claude.ai and Cowork that lets users choose how deeply Claude reasons before responding. Low effort = fast, lightweight responses that conserve rate limits. High effort = slow, deliberate reasoning for complex tasks. It works across all Claude.ai plan tiers.
Is Claude Opus 4.8 available on GitHub Copilot?
Yes. Claude Opus 4.8 became generally available in GitHub Copilot on May 28, 2026 — the same day as the public launch. It is available to all Copilot subscribers (Individual, Business, and Enterprise). No additional configuration is required; select Claude Opus 4.8 from the model picker in VS Code, JetBrains, or the Copilot web interface.
Claude Opus 4.8 represents the clearest signal yet that Anthropic is competing at the very top of the AI market — on benchmarks, on cost, and on developer ecosystem reach. With Dynamic Workflows now in preview and a $965B valuation backing further R&D, the next Opus release is likely less than six weeks away.
Related reading: ChatGPT vs Claude — full comparison | Latest AI News | Best AI Writing Tools 2026
