← Model explorer

Sourced model profile · Current through July 21, 2026

Claude Opus 4.5

A closed-weight Anthropic API model documented for eighteen tasks spanning code generation, mathematical reasoning, vision, and retrieval-augmented generation.

Developed byAnthropic
Released
November 24, 2025

What the evidence says

These facts come from the reviewed Epoch AI catalog artifact and its linked primary record. They describe the model. They do not rank it.

Developer [1]
Anthropic
Released [1]
November 24, 2025
Domains [1]
Language · Multimodal · Vision
Tasks [1]
Code generation · Language modeling/generation · Quantitative reasoning · Search · Visual question answering · Translation · Image captioning · Instruction interpretation · Mathematical reasoning · Visual puzzles · Code autocompletion · Chat · Character recognition (OCR) · Language modeling · Language generation · Text autocompletion · Retrieval-augmented generation · System control
Access [1]
API access
Weights [1]
Closed

Where it may fit

These are evidence-bounded screening suggestions, not performance claims or purchase recommendations.

Code and reasoning evaluation

Consider it for hosted evaluations that span code generation, code autocompletion, and mathematical or quantitative reasoning.

Evidence fields: tasks, access [1]

Document and vision workflows

Consider it when OCR, image captioning, and visual question answering appear in the same workflow as language generation.

Evidence fields: domains, tasks [1]

Limitations

The record documents task scope only. It does not report benchmark results, price, context window, latency, or safety evaluations, and no exact-match Arena rating exists for this record in the pinned snapshot.

Current standing

According to the LMArena snapshot published July 16, 2026, no exact match for this catalog record appears in the style-controlled text category, so this page reports no current rating for Claude Opus 4.5.

Across the boards

No board Model Gauntlet mirrors currently lists this exact model revision under a reviewed identity. This is a coverage gap in the named sources, not a rating.