Claude Opus 4.8
High-stakes reasoning and uncertainty-sensitive work
SWE-Bench Pro leaderlower hallucination exposurestrong OSWorld verified score
Use when silent failure costs more than latency or price.E-GLOBAL-001 / E-GLOBAL-003 / E-GLOBAL-007OCG LabTask-Domain Decision System
Model Reference
Available Models
Current OCG Bench model cards, mapped to task boundaries and global evidence IDs.
Claude Sonnet 4.6
Default high-volume operator
developer-preferred coding workflowsenterprise document readingcost-sensitive triage
Escalate when the task needs cross-source coherence or implicit constraint tracking.E-GLOBAL-005 / E-GLOBAL-009GPT-5.5
Terminal and input-heavy workflow specialist
Terminal-Bench 2.1 leaderCLI-integrated pipelinesinput-volume economics
Avoid as the sole source of truth for unverified extraction or citation-heavy tasks.E-GLOBAL-002 / E-GLOBAL-003Routing Matrix
cross-domain summaryOpus 4.8
Multi-file code requiring cross-state coherence
Legal analysis beyond first-pass triage
Citation-critical document extraction
Financial table reconciliation
Tasks where silent failure is worse than expressed uncertainty
Sonnet 4.6
Single-file debugging under ~300 lines
High-volume document summarization
JSON extraction from labeled inputs
Standard SQL repair
Customer support triage at scale
First-pass legal clause review
GPT-5.5
Terminal-heavy agentic coding
CLI-integrated pipelines
Input-heavy batch tasks where cost is primary constraint