# Claude Opus 5

Claude Opus 5 is an AI model released by Anthropic on Jul 24 2026. It has a 1M context window. Tracked results include 96% on SWE-Bench Verified, 89.5% on SWE-Bench Multilingual and 59.4% on SWE-Bench Multimodal.

## Facts

| Field | Value |
| --- | --- |
| Model | Claude Opus 5 |
| Developer | Anthropic |
| Release date | Friday, Jul 24 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 |
| --- | --- | --- | --- |
| CheatBench | 45.3% | [CheatBench](https://cheatbench.ai/), retrieved 2026-09-25 | Measures how often AI agents try to cheat on difficult assignments, such as reading hidden answers, copying work or manipulating grading. The overall score gives equal weight to ten categories; the sycophancy category measures how far an agent shifts its beliefs toward a user's stated views. Scores describe each model in its tested agent setup, and task success is measured separately. Lower is better. |
| BullshitBench v2 | 73% | [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) | 0.2% | 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) | 1.6% | 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) | 2% | 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 | 4.5% | [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 | 79.2% | 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 | 96% | [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 | 89.5% | 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. |
| SWE-Bench Multimodal | 59.4% | Lab | Real bug reports that arrive with pictures attached — a screenshot, a mockup, a page rendering wrongly — so the AI has to read the image as well as the code to work out what to fix. Higher is better. |
| CursorBench 4.0 | 46.6% | [CursorBench](https://cursor.com/cursorbench), retrieved 2026-09-28 | Cursor's own test of coding agents on ambiguous, multi-file tasks taken from real Cursor sessions — editing, refactoring, investigating a codebase, understanding what the user meant, managing jobs and following a design. Cursor runs each model at several reasoning efforts; each release here carries the score of its best listed effort. Scores aren't comparable with earlier CursorBench versions. Higher is better. |
| DeepSWE 1.1 | 68.8% | 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. (Version 1.1 of the test.) Higher is better. |
| FrontierCode v1.1 (Main) (main split) | 48% | Lab | A set of very hard, frontier-difficulty coding tasks an AI agent has to complete end to end. The score is the share of tasks in the main split it solves. Higher is better. |
| Next.js Evals | 94% | [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. |
| Supabase Evals (with skills) | 92.8% | [Supabase Evals](https://supabase.com/evals) | Supabase's own open benchmark: a coding agent is dropped into a real Supabase project and asked to do real work — set up a schema, fix a broken security policy, debug an Edge Function — and every run is checked against a live Supabase stack. This is the headline number, where the agent has Supabase's own skills loaded, as most people building on Supabase would. The score is the share of scenarios it got right. Higher is better. |
| Supabase Evals (no skills) | 89.9% | [Supabase Evals](https://supabase.com/evals) | The same Supabase scenarios, but with none of Supabase's skills loaded — so it measures what the model already knows about building on Supabase, rather than how well it follows Supabase's supplied instructions. Higher is better. |
| Frontier-Bench v0.1 | 43.3% | 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 | 51.82% | [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-Science 0.1 | 29% | Lab | The same command-line setup as Terminal-Bench, pointed at scientific work: the AI has to drive research tooling and computational workflows through to a result, rather than administer a machine. Version 0.1 is the first release of the task set, and scores run lower than on the general board. Higher is better. |
| BrowseComp | 90.8% | Lab | Can the AI browse the web and track down hard-to-find answers? Higher is better. |
| Humanity's Last Exam (no tools) | 56.3% | 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) | 63.6% | 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-3 | 30.2% | Lab | The third generation of the ARC-AGI series: instead of static puzzles, the AI is dropped into small interactive game-like environments it has never seen and has to figure out the rules and solve them on its own. Higher is better. |
| ARC-AGI-2 | 90.4% | [BenchLM](https://benchlm.ai), retrieved 2026-08-24 | Puzzle-style tests of abstract reasoning and pattern-finding — the kind of thing people find easy but AIs often struggle with. Higher is better. |
| BioMysteryBench (hard) | 49.4% | Lab | Real unsolved-style biology puzzles — the AI has to reason its way to an answer the way a research biologist would. The “hard” split contains the toughest cases. Higher is better. |
| BioMysteryBench (human solved) | 90.1% | Lab | Real biology puzzles that human experts have managed to crack — can the AI reach the same answers? Higher is better. |
| OSWorld 2.0 | 74% | Lab | Can the AI actually operate a computer — clicking, typing, and using real apps — to finish tasks on its own? Version 2.0 is a harder, refreshed task set. Higher is better. |
| AutomationBench | 26.9% | Lab | Tests whether the AI can run real multi-step business workflows — the kind of end-to-end office processes companies want to automate — from start to finish. Higher is better. |
| Harvey's Legal Agent Benchmark (Held-out) | 11.7% | Lab | Harvey's test of whether an AI agent can complete real legal work, scored on a held-out set of tasks the model makers never see — making the numbers harder to game. Higher is better. |
| HealthBench Professional | 59.8% | Lab | Realistic health conversations graded against detailed rubrics written by physicians — can the AI respond the way a careful medical professional would? Higher is better. |
| GDPval-AA v2.1 | 1708 | Lab | economically valuable knowledge work (v2.1, Crowd-BT Elo fit) |
| GDPval-AA v2 | 1861 | Lab | economically valuable knowledge work (v2, re-based Elo) |
| AA-Briefcase | 1685 | Lab | Artificial Analysis agentic office-work eval (Elo) |
| Chartography (with tools) | 83.4% | Lab | A chart-centred test run with tools available to the AI, reported separately from the chart-reading benchmarks above it. Higher is better. |
| threejseval | 1707 | [threejseval](https://threejseval.com/ranking) | Every model gets the same prompt — "the Eiffel Tower", "a glass fishbowl", "a robot arm picking toys into a box" — and builds a 3D scene in Three.js. Real people then see two scenes side by side, names hidden, and vote for the one they prefer. The votes become a chess-style Elo rating on threejseval.com, averaged across all the prompts. It measures whether the scene looks and moves right to a human eye, not whether the code passes a test. Higher is better. |

## About Claude Opus 5

Claude Opus 5 was released on July 24, 2026 as Anthropic's next Opus model. The company positioned it as an everyday option approaching Fable 5's capabilities at half Fable's token price. It became the default model on Claude Max and the highest-capability option on Claude Pro at launch.

Anthropic emphasized improved software engineering, knowledge work and checking results during longer tasks. Its launch evaluations varied effort settings to compare quality with cost. Those comparisons describe specific tasks and configurations; a lower token price and better token efficiency are separate factors in the total cost of completing work.

## Questions and answers

### When was Claude Opus 5 released?

Claude Opus 5 was released by Anthropic on Friday, Jul 24 2026.

### Who made Claude Opus 5?

Claude Opus 5 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 5 cost?

Claude Opus 5 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 5 get?

Claude Opus 5 reports 35 tracked benchmark scores — CheatBench: 45.3%; BullshitBench v2: 73%; Gray Swan IPI (k = 1): 0.2%; Gray Swan IPI (k = 10): 1.6%; Gray Swan IPI (k = 15): 2%; ProgramBench: 4.5%; SWE-Bench Pro: 79.2%; SWE-Bench Verified: 96%; SWE-Bench Multilingual: 89.5%; SWE-Bench Multimodal: 59.4%; CursorBench 4.0: 46.6%; DeepSWE 1.1: 68.8%; FrontierCode v1.1 (Main) (main split): 48%; Next.js Evals: 94%; Supabase Evals (with skills): 92.8%; Supabase Evals (no skills): 89.9%; Frontier-Bench v0.1: 43.3%; Terminal-Bench 4.0: 51.82%; Terminal-Bench-Science 0.1: 29%; BrowseComp: 90.8%; Humanity's Last Exam (no tools): 56.3%; Humanity's Last Exam (with tools): 63.6%; ARC-AGI-3: 30.2%; ARC-AGI-2: 90.4%; BioMysteryBench (hard): 49.4%; BioMysteryBench (human solved): 90.1%; OSWorld 2.0: 74%; AutomationBench: 26.9%; Harvey's Legal Agent Benchmark (Held-out): 11.7%; HealthBench Professional: 59.8%; GDPval-AA v2.1: 1708; GDPval-AA v2: 1861; AA-Briefcase: 1685; Chartography (with tools): 83.4%; threejseval: 1707. 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 Gray Swan IPI (k = 1), Gray Swan IPI (k = 10), SWE-Bench Verified, SWE-Bench Multilingual, SWE-Bench Multimodal, Supabase Evals (with skills), Frontier-Bench v0.1, BioMysteryBench (human solved), Harvey's Legal Agent Benchmark (Held-out) and GDPval-AA v2.

### What is the context window of Claude Opus 5?

Claude Opus 5 has a context window of 1M. Providers may specify separate input and output token limits.

### Is Claude Opus 5 open source?

No. Claude Opus 5 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 5?

Anthropic's previous tracked release was Claude Sonnet 5 on Jun 30 2026, 24 days earlier. It was followed by Claude Fable 5.1 on Sep 1 2026.


---

Canonical page: https://aireleasetracker.com/model/anthropic/claude-opus-5
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.