• featured
claude-opus-4-6 Robot

claude-opus-4-6

Opus 4.6 is Anthropic’s strongest model for coding and long-running professional tasks. It is built for agents that operate across entire workflows rather than single prompts, making it especially effective for large codebases, complex refactors, and multi-step debugging that unfolds over time. The model shows deeper contextual understanding, stronger problem decomposition, and greater reliability on hard engineering tasks than prior generations. Beyond coding, Opus 4.6 excels at sustained knowledge work. It produces near-production-ready documents, plans, and analyses in a single pass, and maintains coherence across very long outputs and extended sessions. This makes it a strong default for tasks that require persistence, judgment, and follow-through, such as technical design, migration planning, and end-to-end project execution. For users upgrading from earlier Opus versions, see our [official migration guide here](https://openrouter.ai/docs/guides/guides/model-migrations/claude-4-6-opus)

Input Balance $4.25 / 1M tokens, Output Cost $21.25 / 1M tokens

Input

No template available.
You can add a prompt template in the admin panel.

Output

Anthropic Claude 4.6 Opus Documentation

Claude 4.6 Opus is designed for sustained, complex knowledge work and multi-step engineering tasks. It features deep reasoning capabilities, exceptional context retention, and strict adherence to architectural constraints, making it highly effective for codebase refactoring, migration planning, and agentic task execution.

Key Capabilities

  • Advanced Problem Decomposition: Automatically breaks down large, ambiguous goals into structured execution plans.
  • Agent-First Architecture: Built to interact seamlessly with external tools, APIs, and file structures across extended sessions.
  • Sustained Knowledge Work: Produces exhaustive, production-ready design documents, technical plans, and codebases in a single pass.
  • Ultra-Long Output Coherence: Maintains stylistic and logical consistency across exceptionally long generation lengths without drifting.
  • High-Fidelity Code Generation: Excels at multi-file debugging, resolving complex circular dependencies, and sweeping framework migrations.

Request Parameters

To interact with this model via an OpenAI-compatible interface, OpenRouter gateway, or direct Anthropic API:

Parameter Type Required Description
model string Yes Use "anthropic/claude-4-6-opus", "claude-opus-4-6", or native "claude-3-7-sonnet-20250219" equivalent.
messages array Yes Array of standard conversation objects (role/content).
max_tokens integer Yes Sets the ceiling for total token generation (including thinking and visible output). Required when thinking is enabled.
thinking object No Configuration for the model's reasoning capabilities. See Thinking Mode Configuration below.
output_config object No Controls reasoning effort and depth when adaptive thinking is active.
temperature float No Recommended: 0.0 or 0.1 for precise engineering tasks. Note: Must be 1.0 if thinking is enabled.
system string No Specific system instruction string (if passed at root level).

Thinking Mode Configuration

Claude 4.6 Opus utilizes an updated Adaptive Thinking mechanism. The older "type": "enabled" and budget_tokens parameters are deprecated for this model. To control the model's deep reasoning, use the new adaptive configuration combined with output_config.

Important Rules for Thinking Mode:

  1. Temperature Constraint: If thinking is included in the request, temperature must be set to 1.0 (or omitted, as it defaults to 1.0). Setting it to any other value will result in a 400 Bad Request error.
  2. Token Ceiling: max_tokens must be large enough to accommodate both the hidden thinking process and the final visible response (e.g., 16000).

Request Example (JSON)

json Copy 
{
  "model": "claude-opus-4-6",
  "max_tokens": 16000,
  "temperature": 1.0,
  "thinking": {
    "type": "adaptive"
  },
  "output_config": {
    "effort": "high"
  },
  "messages": [
    {
      "role": "user",
      "content": "Analyze this legacy architecture and design a zero-downtime migration plan to a microservices mesh."
    }
  ]
}

Parameter Breakdown

thinking (object)

  • type (string, Required): Must be set to "adaptive".

    Note: Setting this to "enabled" on Claude 4.6 Opus will trigger a deprecation warning or a 400 error depending on the gateway version.

output_config (object)

  • effort (string, Optional): Controls the depth and duration of the thinking phase.
    • "low": Fast responses, minimized thinking for simpler sub-tasks.
    • "medium": Balanced speed and reasoning depth.
    • "high": Standard deep reasoning for complex engineering (Default).
    • "max": Maximum reasoning capacity for highly experimental or complex architectural problems.

Unlock the most affordable AI hosting

Run models at scale with our fully managed GPU infrastructure, delivering enterprise-grade uptime at the industry's best rates.

Contact Sales