Model introduction
Claude Opus 5 is Anthropic's July 2026 high-capability model for complex agentic coding, code review, enterprise knowledge work, and tasks that require sustained autonomous execution.
Model capabilities
Specifications come from the cited model material. Catalog tags help identify access features; use the source documentation for exact limits.
Verified specifications
- Official positioning
- Complex agentic coding and enterprise work
- Release date
- 2026-07-24
- Knowledge cutoff
- 2026-05
- Context window
- 1,000,000 tokens
- Maximum synchronous output
- 128,000 tokens
- Input / output
- Text, image / Text
- Reasoning levels
- low, medium, high, xhigh, max; high by default
- Prompt-cache minimum
- 512 tokens
- Verified capabilities
- Functions, structured outputs, context editing, memory, code execution, programmatic tools, compaction
- Platform availability
- Claude API, AWS Bedrock, Google Vertex AI, Microsoft Foundry
Sources: Anthropic Claude Opus 5 update notes
Independent capability evaluation
Only public measurements matched to the exact model ID and reasoning profile are shown; nearby variants are not substituted.
Intelligence index
60
Cohort rank #2 / 190
Output speed
52.4 tok/s
Cohort rank #117 / 190
First-token latency
33.84s
Adaptive reasoning, xhigh effort
The xhigh profile ranks near the top for composite capability, but its measured output speed and first-token delay favor valuable complex work over low-latency interaction.
Evaluation profile: Adaptive reasoning, xhigh effort. Artificial Analysis · Methodology
Model pricing and access
Prices are read directly from the MoleAPI console catalog and shown by current billing group and context tier.
Your final charge follows the account group shown in the console.
Live pricing source
Pricing table
USD / 1M tokensStandardx1defaultInput $5 · Output $25
Input
$5 / 1M tokens
Output
$25 / 1M tokens
Cache read
$0.5 / 1M tokens
Cache write
$6.25 / 1M tokens
Discountx0.8discountInput $4 · Output $20
Input
$4 / 1M tokens
Output
$20 / 1M tokens
Cache read
$0.4 / 1M tokens
Cache write
$5 / 1M tokens
Relayx0.3relayInput $1.5 · Output $7.5
Input
$1.5 / 1M tokens
Output
$7.5 / 1M tokens
Cache read
$0.15 / 1M tokens
Cache write
$1.875 / 1M tokens
Access protocols
Available billing groups: Standard, Discount, Relay
- anthropic
- POST
- /v1/messages
- openai
- POST
- /v1/chat/completions
Complete model guide
Prepared from vendor sources and independently reviewed. · 2026-07-28
Claude Opus 5 is Anthropic’s model for complex agentic coding and enterprise work. Anthropic describes it as a generation that improves on Opus 4.8 in deep reasoning, agentic and long-horizon tasks, and test-time compute scaling; it supports text and image input, text output, and has adaptive thinking enabled by default. [1][2]
Model Positioning and Core Capabilities
Claude Opus 5 is designed for complex coding and enterprise work that requires execution across files, tools, and multiple stages. Its reasoning depth can be controlled with the effort parameter, with adaptive thinking as the default behavior. [1]
Anthropic highlights its capabilities in deep reasoning, long-horizon agentic tasks, and test-time compute scaling. The model provides five reasoning effort levels—low, medium, high, xhigh, and max—with high as the default. [1]
Capabilities and Specifications
Claude Opus 5 provides a 1 million token context window and a maximum output of 128,000 tokens, supporting text and image input and text output. [1][2]
| Specification | Value |
|---|---|
| Context window | 1,000,000 tokens [1] |
| Maximum output | 128,000 tokens [1] |
| Input modalities | Text, images [2] |
| Output modalities | Text [2] |
| Reasoning mode | Adaptive thinking enabled by default [1] |
| Reasoning effort levels | low, medium, high, xhigh, max; default: high [1] |
| Knowledge cutoff | May 2026 [1] |
max_tokens constrains the total output budget for both thinking tokens and visible text responses, rather than only limiting the final response length. [1]
The model overview lists Claude Opus 5 as a Claude model with multilingual and vision capabilities. [2]
Use Cases
Claude Opus 5 covers software engineering tasks such as multi-file feature development, large-scale refactoring, code review, and bug investigation. Anthropic’s prompting guide also uses it for end-to-end task execution, long-context tool use, and multi-agent collaboration. [3]
For enterprise knowledge work, official listed applications include understanding charts, documents, and diagrams; processing complex multi-sheet spreadsheets; and generating presentations. [3]
Important Limitations
Thinking cannot be disabled for Claude Opus 5 at the xhigh or max reasoning effort levels; combining either level with thinking: {"type":"disabled"} causes the API to return a 400 error. [1]
Disabling thinking applies only to high and lower effort levels. Anthropic also notes that when thinking is disabled, the model may occasionally write tool calls into text output rather than tool_use blocks, or include internal XML tags in visible responses. [1]
Sources
[1] Anthropic Claude Opus 5 更新说明 [2] Anthropic Claude 模型总览 [3] Claude Opus 5 提示词工程指南
Code examples
These examples use MoleAPI's OpenAI-compatible endpoint and run after you replace the API key.
curl https://api.moleapi.com/v1/chat/completions \
-H "Authorization: Bearer $MOLEAPI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-opus-5",
"messages": [{"role": "user", "content": "Hello"}]
}'Frequently asked questions
How is claude-opus-5 priced?
This page reads prices from the MoleAPI console API and updates with model prices, context tiers, and account groups.
How can I access claude-opus-5?
Protocols and endpoints come from the supported_endpoint_types field in the MoleAPI model catalog.
How do I switch an existing project to claude-opus-5?
Keep the MoleAPI API address and key, replace the model parameter with the model ID on this page, then check protocol-specific parameter differences.
Where do the claude-opus-5 model details come from?
Capabilities and limitations are checked against Anthropic and the other cited pages. Pricing and available protocols come only from the MoleAPI console.