Anthropic · The most demanding information tasks
Claude Opus 5
A choice for complex analysis, long tasks involving multiple steps and outputs where quality matters more than speed and cost.
- Costs
- High · $5 per million input tokens and $25 per million output tokens; a token is a small piece of text
- Speed
- Rather slow at the most thorough setting
- Availability
- Online service and apps from Anthropic
- Input length
- Very long input · up to 1 million tokens (small pieces of text)
- Inputs
- Text and image inputs
Data checked . Release date unverified.
Indicative capability profile
Where the model is strong
The five levels are our clear summary of the results below. They are not a ranking that applies to every task.
Ideal use
When to choose it
- Complex research and supporting information for decisions
- Long tasks the model solves in several steps
- Advanced coding and independent code review
Usage boundaries
When to choose another model
- Routine sorting and simple transcriptions
- Interactive tasks requiring an immediate response
Traceable supporting sources
Data and measurement sources
Distinguish between the manufacturer's documentation and the results of a specific test. Measurements also depend on the settings and the task set used.
Intelligence Index v4.1
61 pointsAt the top of the aggregate measurement as of 24 July 2026.
Open original source ↗AA-Briefcase
1720 Elo pointsA higher number is better. The model led in working with information; its tools were also tested.
Open original source ↗Terminal-Bench v2.1
89 %Very good result in coding and terminal work.
Open original source ↗Trends over time
Related events from AI Radar
Astra and Claude Opus 5 models help decipher two long-unsolved Enigma messages
The difference in approach is notable: the Astra model worked with minimal input and sourced the context itself, while the Claude Opus 5 model needed more concrete guidance from a human – so when assigning similar AI research tasks, it's worth considering how much initiative to expect from the model.
The case shows that current state-of-the-art models can autonomously carry out multi-stage archival research, build simulations, and solve tasks without a clearly defined procedure – a relevant signal for companies considering deploying AI agents on complex analytical and research tasks where no ready-made methodology exists.
Pixel-art kakapo animation created with the Claude Opus 5.5 model and converted into video using Claude Code and Playwright
Anyone creating presentations can use this approach to quickly produce a short animated video from HTML5 canvas content without using a video editor or graphics tool.
Anthropic challenges US approach in the Claude Fable 5 and Mythos 5 export control case, responds with jailbreak severity framework
New: Export control issued due to concerns about a jailbreak and the model's cyber capabilities; Anthropic challenged the legality of the export control and the lack of a clear process; Anthropic responded by publishing a jailbreak severity framework
Security researchers and developers now have access to a formal Anthropic program and framework for reporting and assessing the severity of jailbreaks affecting Claude models.
Companies dependent on the Claude Fable 5 or Mythos 5 models face the risk that US export controls could temporarily restrict their access to the model without a clearly defined process, which is relevant for supply chain and geopolitical risk planning.
Anthropic pushes for mandatory AI system safety audits in the US
The proposed legislation (e.g., the Massachusetts bond bill) would introduce an obligation to disclose safety frameworks and undergo independent third-party audits for companies developing or deploying AI systems, which would increase compliance costs and risk-documentation requirements.
Next step
Compare the model using your actual task.
A benchmark narrows the selection. A short trial on your data determines the choice.