Skip to content
Longterm Wiki
Index
Ai Model·claude-opus-4-6·Record·Profile

AI Model: Claude Opus 4.6 (anthropic)

Verdictpartial92%
1 check · 4/30/2026

1 → partial

Our claim

entire record
Entity Type
ai-model
Title
Claude Opus 4.6
Description
Claude Opus 4.6 was released February 5, 2026. Introduced adaptive thinking with four effort levels, agent teams for splitting tasks, and 1M-token context window (beta). Scored 68.8% on ARC-AGI-2 (83% improvement over Opus 4.5), 65.4% on Terminal-Bench 2.0, and 72.7% on OSWorld. expandClaude Opus 4.6 was released February 5, 2026. Introduced adaptive thinking with four effort levels, agent teams for splitting tasks, and 1M-token context window (beta). Scored 68.8% on ARC-AGI-2 (83% improvement over Opus 4.5), 65.4% on Terminal-Bench 2.0, and 72.7% on OSWorld. Priced at \$5/\$25 per million tokens.
Tags
claude,claude-4,adaptive-thinking,agent-teams
Last Updated
February 2026
Related Entries
[object Object],[object Object]
Metadata
[object Object]

Source evidence

1 src · 1 check
partial92%Haiku 4.5 · 4/30/2026

NoteThe source confirms most fields: title (Claude Opus 4.6), developer (Anthropic), releaseDate (Feb 5, 2026), inputPrice ($5 per million tokens), outputPrice ($25 per million tokens), and contextWindow (1M tokens). However, the safetyLevel field 'ASL-3' is not mentioned anywhere in the source text. The source discusses safety extensively but uses different terminology ('low rates of misaligned behavior', 'as good as, or better than, any other frontier model', references to 'system card'). The safety level designation 'ASL-3' appears to be an external classification system not referenced in the source material. This makes the safety level claim unverifiable from this source, resulting in a 'partial' verdict since 5 of 6 key fields are confirmed but 1 cannot be verified.

Case № claude-opus-4-6Filed 4/30/2026Confidence 92%