Claude Opus 4.6
Anthropic's flagship language model from the Claude 4 family, released February 5, 2026, featuring a 1M token context window, adaptive reasoning, and advanced agentic capabilities.
Technical specification
Modalities
Capabilities
12Reasoning★
Reasoning
Multi-step reasoning★
Reasoning
Long context★
Reasoning
Coding★
Coding
Function Calling
Planning
Structured output★
Structured gen.
Image understanding★
Vision
Chart understanding
Vision
OCR★
Vision
Multilingual★
Language
Planning★
Planning
Streaming output
Reasoning
Standard API pricing: $5 per million input tokens, $25 per million output tokens. Full 1M token context window available at standard pricing (no long-context surcharge). Prompt caching: write $6.25/MTok (TTL 5 min) or $10/MTok (TTL 1 hr); cache read $0.50/MTok. Batch API: 50% discount. Fast mode (research preview): $30 / $150 per million tokens. In Batch API, up to 300K output tokens available using the output-300k-2026-03-24 header. US-only inference (inference_geo): 1.1× multiplier. Pricing unchanged from Claude Opus 4.5.
INPUT
$5.0000 / per 1M tokens
OUTPUT
$25.0000 / per 1M tokens
CACHE
$0.5000 / per 1M tokens
TOTAL
for 10K tokens
Standard synchronous Messages API. Cache writes: $6.25/MTok (TTL 5 min) or $10/MTok (TTL 1 hr). Max 128K output tokens synchronously; up to 300K with the beta header Batch API.
Batch API (50% discount); asynchronous processing. Supports up to 300K output tokens with the output-300k-2026-03-24 header.
Fast mode (research preview) — available exclusively for Opus 4.6. Enables up to 2.5× faster output token generation. Not available in the Batch API.
Security and enterprise
Claude Opus 4.6 demonstrates a safety profile at least as strong as Opus 4.5, with low rates of policy-violating behavior. On Amazon Bedrock, a zero operator access policy applies. Data residency options are available via the inference_geo parameter (1.1× multiplier for US-only inference). Enterprise security and compliance details: trust.anthropic.com.
Technical information
Full safety evaluation is available in the Claude Opus 4.6 system card at anthropic.com/claude-opus-4-6-system-card. Opus 4.6 achieves the lowest over-refusal rate among recent Claude models. Enhanced planning capabilities may theoretically increase the potential for obfuscation in cases of misalignment — this issue is discussed in the system card.