gpt.college

Anthropic · Claude · Reviewed 2026-06-11

Claude Opus 4.6

Anthropic's Opus 4.6 release, positioned around long-running agentic coding, computer-use, and professional workflows.

All models
Overall · 0–100—
Status
active
Released
—
Context
Not disclosed
Parameters
Undisclosed
Inputs
Text · Image · File
Type
multimodal
Max output
— tokens
API model ID
—
Reasoning
—
Knowledge cutoff
—
Training cutoff
—
Weights / license
—
Price / 1M tokens
—
Cached input
—
Strong official SWE-bench Verified resultAgentic workflow positioningLong-context workflow emphasis
CodingReasoningResearchMultimodalAgentic

Independent benchmark scores

No snapshot score

Coding

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Reasoning

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Research

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Multimodal

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Agentic

—

No comparable score for this task lens in the selected independent snapshot.

Raw benchmark results

Source-backed rows with the original evaluation context preserved.

2 results
BenchmarkScoreSourceEvaluation notes
Humanity's Last Exam (HLE)Checked 2026-06-12 36.24% Benchmark project Scale Labs HLE Text Only score for claude-opus-4-6-thinking-max, covering the 86% text-only HLE subset.
SWE-bench VerifiedChecked 2026-06-11 81.42% Anthropic Anthropic reports this SWE-bench Verified score averaged over 25 trials with a prompt modification.

Strengths

  • Strong official SWE-bench Verified result
  • Agentic workflow positioning
  • Long-context workflow emphasis

Limitations

  • Some benchmark notes include repeated trials or prompt modifications; compare harnesses carefully

Model evidence · official · Checked 2026-06-10

Claude models overview

Use for current model family and capability notes; verify before publishing claims.

Model evidence · official · Checked 2026-06-11

Introducing Claude Opus 4.6

Official Anthropic release page. Use benchmark notes carefully because some reported scores use repeated trials or prompt modifications.

Task lenses preserve raw scores. Radar axes use score / current leader × 100; overall is the equal-weight average of all five indices. Missing evidence is not a zero or a low rank.