Claude Opus 4.8

Name: Claude Opus 4.8
Author: Anthropic

Multimodal

Anthropic

Claude Opus 4.8 is Anthropic's upgrade to Opus 4.7 and its most capable general-access model at release, with improvements across software engineering, agentic tool use, reasoning, computer use, and knowledge-work benchmarks while shipping at the same price ($5/$25 per million input/output tokens).

Key Specifications

Parameters

Context

1.0M

Release Date

May 28, 2026

Average Score

72.8%

API Documentation Research Paper Results Blog

Timeline

Key dates in the model's history

Announcement

May 28, 2026

Last Update

August 27, 2026

Technical Specifications

Parameters

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$5.00

Output (per 1M tokens)

$25.00

Max Input Tokens

1.0M

Max Output Tokens

128.0K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

Standard harness with thinking blocks included in sampling, averaged over 5 trials. • Self-reported

88.6%

Reasoning

Logical reasoning and analysis

GPQA

GPQA Diamond. Averaged over 25 trials. • Self-reported

93.6%

Other Tests

Specialized benchmarks

BrowseComp

Single-agent with web search, web fetch, code execution, adaptive thinking at max effort, 10M-token limit, context compaction at 200k tokens. Multi-agent (orchestrator + blocking subagents) reaches 88.5%. • Self-reported

84.3%

CharXiv-R

CharXiv Reasoning. Adaptive thinking at max effort with Python tools: 89.9%. Without tools: 80.5%. 1,000 validation questions averaged over 5 runs. • Self-reported

89.9%

CyberGym

Targeted vulnerability reproduction (pass@1) over 1,507 tasks without deployed safeguards. With Tier-3 safeguards: 1.0%. • Self-reported

78.8%

DeepSearchQA

F1 score. Web search, web fetch, code execution, max reasoning effort, adaptive thinking, 1M token budget. • Self-reported

93.1%

DeepSWE 1.1

DeepSWE v1.1 leaderboard, mini-swe-agent harness, max effort; Pass@1 59% ± 2%. • Self-reported

59.0%

Finance Agent

Agentic financial analysis (Finance Agent v2). Evaluated by Vals AI with adaptive thinking at max effort. • Self-reported

53.9%

Finance Agent v2

• Self-reported

53.9%

FrontierCode 1.1

FrontierCode 1.1 current leaderboard; mergeability score at max effort: 46.5%. • Self-reported

46.5%

FrontierSWE

Claude Code • Self-reported

75.0%

Graphwalks BFS >128k

F1 score on the 1M-token subset, averaged over 5 trials. 256K subset: 85.9%. • Self-reported

68.1%

Graphwalks parents >128k

F1 score on the 1M-token subset, averaged over 5 trials. 256K subset: 99.3%. • Self-reported

83.3%

HealthBench Professional

Length-adjusted score with adaptive thinking at max effort. Claude Sonnet 4.6 as grader, averaged over 5 trials. • Self-reported

55.8%

Humanity's Last Exam

Multidisciplinary reasoning at max effort. With tools (web search, web fetch, code execution): 57.9%. Without tools: 49.8%. • Self-reported

57.9%

Include

Multilingual regional-knowledge benchmark covering 44 languages. Adaptive thinking enabled with structured JSON output. • Self-reported

87.6%

LiveBench

2026-01-08, Thinking xHigh Effort • Self-reported

77.2%

MCP Atlas

Real-world MCP tool use across production-like servers. Evaluated by Scale AI (April 2026 config: 100 tool-call budget). Mean claim coverage of 86.2%. • Self-reported

82.2%

OfficeQA Pro

Internal agentic harness with extracted-text documents and code-execution tools. OfficeQA full set: 77.6%. • Self-reported

66.2%

OSWorld-Verified

Agentic computer use evaluation. 361 tasks at 100 steps, 1080p resolution, adaptive thinking at max effort. Pass@1 averaged over 5 seeds. Anthropic updated harness methodology (zoom-tool fix, 128K max tokens per turn). • Self-reported

83.4%

ScreenSpot Pro

GUI grounding. Adaptive thinking at max effort with Python tools: 87.9%. Without tools: 82.3%. Averaged over 5 runs. • Self-reported

87.9%

SWE-bench Multilingual

300 problems across 9 programming languages. • Self-reported

84.4%

SWE-Bench Multimodal

Visual context (screenshots, design mockups) added to issue descriptions. Internal harness. • Self-reported

38.4%

SWE-Bench Pro

Agentic coding evaluation. Standard harness, adaptive thinking at max effort, averaged over 5 trials. • Self-reported

69.2%

Terminal-Bench 2.0

Terminal-Bench 2.1 with the Terminus-2 public harness on Daytona, high effort, mean reward over 5 attempts per task (89 tasks, 445 trials). • Self-reported

74.6%

Toolathlon

Pass@1 averaged over 3 trials across all 108 tasks. Internal harness with adaptive thinking at max effort. Pass@3: 67.6%. • Self-reported

59.9%

License & Metadata

License

proprietary

Announcement Date

May 28, 2026

Last Updated

August 27, 2026

Similar Models

All Models

Claude Opus 4.6

Anthropic

Best score:1.0 (TAU)

Released:Feb 2026

Price:$5.00/1M tokens

Claude Sonnet 4.6

Anthropic

Best score:0.9 (GPQA)

Released:Feb 2026

Price:$3.00/1M tokens

Claude Sonnet 4.5

Anthropic

Best score:0.9 (TAU)

Released:Sep 2025

Price:$3.00/1M tokens

Claude Opus 4.7

Anthropic

Best score:0.9 (GPQA)

Released:Apr 2026

Price:$5.00/1M tokens

Claude Opus 4.5

Anthropic

Best score:0.9 (TAU)

Released:Nov 2025

Price:$5.00/1M tokens

Claude Sonnet 5

Anthropic

Released:Jun 2026

Price:$2.00/1M tokens

Claude Opus 5

Anthropic

Released:Jul 2026

Price:$5.00/1M tokens

Claude Fable 5

Anthropic

Released:Jun 2026

Price:$10.00/1M tokens

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.