# Claude Opus 4.8

Claude Opus 4.8 is an AI model released by Anthropic on May 28 2026. It has a 1M context window. Tracked results include 95% on BullshitBench v2, 74% on BU Bench and 88.6% on SWE-Bench Verified.

## Facts

| Field | Value |
| --- | --- |
| Model | Claude Opus 4.8 |
| Developer | Anthropic |
| Release date | Thursday, May 28 2026 |
| Licensing | Closed |
| Context window | 1M |

## API pricing

All rates in USD per 1,000,000 tokens, pay-as-you-go.

| Tier | Input | Cached input | 5 min cache write | 1 hr cache write | Output |
| --- | --- | --- | --- | --- | --- |
| All documented contexts | $5.00 | $0.50 | $6.25 | $10.00 | $25.00 |

Verified October 19, 2026 against the first-party source: https://platform.claude.com/docs/en/about-claude/pricing

## Tracked benchmark scores

| Benchmark | Score | Source | What it measures |
| --- | --- | --- | --- |
| BullshitBench v2 | 95% | [BullshitBench](https://github.com/petergpt/bullshit-benchmark) | Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. |
| Gray Swan IPI (k = 1) | 0.5% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. Gray Swan's indirect prompt injection benchmark measures how often such an attack succeeds when the attacker gets a single try. Lower is better. |
| Gray Swan IPI (k = 10) | 4.1% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 10 tries and counts an attack as successful if any of them works. Lower is better. |
| Gray Swan IPI (k = 15) | 5.5% | Lab | Attackers hide malicious instructions inside content the AI reads — a web page, an email, a document — and try to hijack what it does. This variant gives the attacker 15 tries and counts an attack as successful if any of them works. Lower is better. |
| ProgramBench | 0% | [ProgramBench](https://programbench.com/), retrieved 2026-09-11 | The AI receives a working program and its documentation, then builds a replacement from scratch without the original source code, internet access or decompilation. The score is the percentage of 200 programs that pass every behavioral test. We record each model's best published mini-SWE-agent result, including higher reasoning efforts where available. Partial test-pass rates and almost-solved programs do not count toward this score. Equal scores share a rank here; the official board also uses partial progress to break ties. Higher is better. |
| SWE-Bench Pro | 69.2% | Lab | Can the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. |
| SWE-Bench Verified | 88.6% | [BenchLM](https://benchlm.ai), retrieved 2026-08-24 | Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better. |
| SWE-Bench Multilingual | 84.4% | Lab | Like SWE-Bench, but the coding problems span many programming languages, not just one. Tests how broadly the AI can code. Higher is better. |
| DeepSWE 1.0 | 55.8% | Lab | Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. Higher is better. |
| Next.js Evals | 68% | [Next.js Evals](https://nextjs.org/evals) | Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better. |
| Frontier-Bench v0.1 | 21.1% | Lab | A hard, ever-evolving set of real computer tasks — coding, system administration, data work, and more — that an AI agent has to complete on its own. Run by the Harbor / Laude Institute team as the successor to Terminal-Bench (v0.1 is the first release of the task set). The score is the share of tasks solved. Higher is better. |
| Terminal-Bench 4.0 | 23.64% | [Terminal-Bench](https://github.com/harbor-framework/terminal-bench), retrieved 2026-08-30 | Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Version 4.0 recalibrated how much time, CPU and memory each task gets, removed eight tasks and fixed nineteen, so fewer runs fail for reasons that have nothing to do with the model. Scores are not comparable with earlier versions. Higher is better. |
| Terminal-Bench 2.1 | 74.6% | Lab | Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better. |
| BU Bench | 74% | Lab | Can the AI drive a real web browser to finish tasks — clicking, filling forms, and navigating sites the way a person would? Run by Browser Use on their BU Bench task set. Higher is better. |
| BrowseComp | 84.3% | [BenchLM](https://benchlm.ai), retrieved 2026-07-11 | Can the AI browse the web and track down hard-to-find answers? Higher is better. |
| Humanity's Last Exam (no tools) | 49.8% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. |
| Humanity's Last Exam (with tools) | 57.9% | Lab | Humanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. |
| ARC-AGI-2 | 72.08% | [BenchLM](https://benchlm.ai), retrieved 2026-08-31 | Puzzle-style tests of abstract reasoning and pattern-finding — the kind of thing people find easy but AIs often struggle with. Higher is better. |
| OSWorld-Verified | 83.4% | Lab | Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Higher is better. |
| Finance Agent v2 | 53.9% | Lab | Tests the AI on real financial-analysis work, like digging through reports and making sound decisions. Higher is better. |
| Harvey's Legal Agent Benchmark | 9.58% | Lab | Harvey's test of whether an AI agent can complete real legal work — drafting and reviewing documents, working with spreadsheets and presentations, and navigating files the way a lawyer's assistant would. Higher is better. |
| TaxEval v2 | 75.63% | Lab | A set of real tax questions created by Vals AI — can the AI give accurate answers about tax rules and filings? Higher is better. |
| MedScribe | 85.75% | Lab | Can the AI support doctors with their administrative work, like notes and paperwork? Created by Vals AI. Higher is better. |
| GDPval-AA | 1890 | Lab | Measures how well the AI does economically valuable knowledge work, judged against human experts. Shown as a rating (like a chess Elo) — higher is better. |
| GDPval-AA v2 | 1600 | Lab | economically valuable knowledge work (v2, re-based Elo) |

## About Claude Opus 4.8

Claude Opus 4.8 was released on May 28, 2026 as an update to Opus 4.7, with the same regular API token prices. Anthropic emphasized better judgment during agent tasks, improved coding and a greater tendency to flag uncertainty or check its work. The company published evaluation and safety details in a separate system card.

The launch also introduced effort controls in Claude's apps and a research preview of dynamic workflows in Claude Code. Those workflows let the coding system organize work across parallel subagents. They were product capabilities around the model, rather than evidence that a single request always used multiple agents or that every benchmark score was an independent measurement.

## Questions and answers

### When was Claude Opus 4.8 released?

Claude Opus 4.8 was released by Anthropic on Thursday, May 28 2026.

### Who made Claude Opus 4.8?

Claude Opus 4.8 was built by Anthropic. AI safety company building the Claude family of models. Founded in 2021 by former OpenAI researchers.

### How much does Claude Opus 4.8 cost?

Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens through the Anthropic API. Cached input is $0.50 per million tokens. Rates are pay-as-you-go API prices verified against Anthropic's published pricing on October 19, 2026.

### What benchmark scores did Claude Opus 4.8 get?

Claude Opus 4.8 reports 25 tracked benchmark scores — BullshitBench v2: 95%; Gray Swan IPI (k = 1): 0.5%; Gray Swan IPI (k = 10): 4.1%; Gray Swan IPI (k = 15): 5.5%; ProgramBench: 0%; SWE-Bench Pro: 69.2%; SWE-Bench Verified: 88.6%; SWE-Bench Multilingual: 84.4%; DeepSWE 1.0: 55.8%; Next.js Evals: 68%; Frontier-Bench v0.1: 21.1%; Terminal-Bench 4.0: 23.64%; Terminal-Bench 2.1: 74.6%; BU Bench: 74%; BrowseComp: 84.3%; Humanity's Last Exam (no tools): 49.8%; Humanity's Last Exam (with tools): 57.9%; ARC-AGI-2: 72.08%; OSWorld-Verified: 83.4%; Finance Agent v2: 53.9%; Harvey's Legal Agent Benchmark: 9.58%; TaxEval v2: 75.63%; MedScribe: 85.75%; GDPval-AA: 1890; GDPval-AA v2: 1600. Tracked scores may come from lab reports or independent benchmarks; source details accompany the benchmark data. It holds the best score among all models tracked here on BullshitBench v2 and BU Bench.

### What is the context window of Claude Opus 4.8?

Claude Opus 4.8 has a context window of 1M. Providers may specify separate input and output token limits.

### Is Claude Opus 4.8 open source?

No. Claude Opus 4.8 is a closed model. The weights are not published — it is available only through the provider's own API, apps, or partner platforms.

### What came before and after Claude Opus 4.8?

Anthropic's previous tracked release was Claude Opus 4.7 on Apr 16 2026, 42 days earlier. It was followed by Claude Fable 5 on Jun 9 2026.


---

Canonical page: https://aireleasetracker.com/model/anthropic/claude-opus-4.8
Site index: https://aireleasetracker.com/llms.txt
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.