Unverified50% confidenceBenchmarkExact time

In the Tīrtha Bench shared comparison, the highest accuracy is 95.4%, achieved jointly by Grok 4.7 and Claude Opus 5.5.

1

Sources

50%

Confidence

Short-term (~14 days)

Relevance

10/19/2026

First Seen

Valid until: 11/2/2026

Sources

Tīrtha Bench: The Data Programmers Should Look At When Choosing AI Models

Related Entities

Tīrtha Benchproduct Grok 4.7model Claude Opus 5.5model

Related Claims

UnverifiedGrok 4.5's performance on Tīrtha Bench was 93.4% accuracy, a 98.3% trap pass rate, 1.85 seconds latency, and $0.0015 per correct answer.77% similar UnverifiedIn the MemoryBench test, Muse Wonder achieved 85.5% accuracy, exceeding Claude Opus 5.5's 81.5% and GPT-6 Astra's 81.4%.74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/1064717

API

curl https://kongchang.com/api/v1/knowledge/claims/1064717

MCP

get_claim(id=1064717)