LANGUAGE MODEL Dec 2025 Anthropic Last updated: Apr 29, 2026

Claude Opus 4.6

1M Context, Sustained Multi-Hour Agency

Anthropic's December 2025 Opus refresh, with the headline addition of a 1-million-token context variant. The agentic headline: sustained autonomous coding sessions extending beyond 12 hours uninterrupted (vs Opus 4's ~7 hour benchmark from May 2025). The model behind most "AI did multi-day project autonomously" demos in late 2025.

Try Claude API Docs ↗

Intelligence

Top 10%

Speed

Slow

52 tok/s output

Cost

High

$6.25 in / $25.00 out

Context

Up to 1,000,000 tokens

How are Intelligence, Speed & Cost bucketed?

Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).

Intelligence

Top 1%≤ 1%
Top 5%≤ 5%
Top 10%≤ 10%
Good≤ 25%
Medium≤ 50%
Below avg> 50%

Speed

Top 1%≥ 345 tok/s
Top 5%≥ 237 tok/s
Top 10%≥ 196 tok/s
Good≥ 146 tok/s
Medium≥ 90 tok/s
Slow< 90 tok/s

Cost

Freeopen weights · self-host
Low< $1 / M out
Moderate$1–5 / M out
High≥ $5 / M out

Official ↗ Artificial Analysis ↗

Why it matters

Opus 4.6 is the model that this entire codebase is likely being authored by — its capability profile is what made "build a content site with 50+ MDX nodes plus dual orientation graph + light theme + Civ-style layout in one session" a tractable conversation rather than a multi-week project.

Core Capabilities

Long Documents

Handles entire codebases, books, and multi-doc RAG.

Multimodal

Combines text, vision, and audio in one model.

Generative

Produces images, video, audio, or other media.

Agent Workflows

Built for tool use and autonomous tasks.

Context Window

1M tokens

≈ entire codebase

4k Chat 聊天

32k Long docs 长文档

128k Books 整本书

400k Multi-doc 多文档

1M This model 本模型

10M

Availability

API

Available

Product / App

Available

Open Source

Not released

Enterprise

Available

Pricing Model

Pay per token

Input and output billed separately.

Pay-per-token

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model

Reasoning

AA Intelligence Index · scaled to 10

1.7

5.6

7.6

Coding

SciCode · scaled to 10

1.8

4.3

5.2

Agentic tasks

Terminal-Bench Hard · scaled to 10

0.2

3.6

4.6

Context / memory

Context window size · log-scaled

6.0

9.0

Cost efficiency

Input price ($/M tokens) · cheaper scores higher

6.2

10.0

3.2

Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

First Opus with 1M-token context window — the long-context gap to Gemini 2.5 Pro finally closes
Tops Terminal-Bench 2.0 (65.4%), OSWorld (72.7%), BrowseComp (84.0%), Finance Agent (60.7%)
53.1% on Humanity's Last Exam with tools — beats every other frontier model at release
ARC-AGI-2 score of 68.8% — biggest single-jump on novel-problem-solving in any model so far
Adaptive thinking + conversation compaction make multi-hour sessions much more stable
SWE-bench Verified 80.8% (vs 80.9% on Opus 4.5) — coding plateaued; the gains are in agents and reasoning