FOLLOW A THREAD

基准测试

Connect scattered updates into a story you can understand.

ProjectsNews10/15, 00:53

xAI launches Grok Voice 3.0, saying it ranks Speech-to-Speech Index first

Beating AI news brief says xAI launched Grok Voice 3.0 and made it available through Voice Agent API, maintaining pricing at 0.08 USD。

MarketNews10/13, 08:29

Meta released the Muse Wonder series AI models,Meta stock closed higher 3.38%

Meta released two models, Muse Wonder and Muse Wonder Full,AI saying that they matched or surpassed the performance of some leading competitors in certain benchmarks. After the news was announced,Meta U.S. shares closed higher 3.38%。

# Breaking News # Meta # AI Model # benchmark testing # Muse assistant

ProjectsNews10/10, 18:54

Pine launches a cloud computer for AI Agent; benchmark performance and complete task completion rate diverge

Pine AI launched the invitation-only beta of Pine Computer, saying that it improves Agent execution efficiency through system-level change feedback. Tests cited in the report showed that its checkpoint score led the field, but its complete task completion rate was lower than those of Claude Code and Codex.

# Industry Insights # AI Agent # cloud computer # benchmark testing # Developer tools

ProjectsNews10/10, 09:39

Pine AI announced the launch of a cloud computer for AI agents called Pine Computer

Pine AI announced the launch of Pine Computer, a cloud computer for AI agents. Developers can SDK call it to perform tasks across websites, files, and business software. The product is currently in a private developer testing phase.

# Technology Developments # AI Agents # cloud computing # Developer tools # benchmark testing

MarketNews10/05, 12:26

Report says performance gap between top Chinese and U.S.AI models narrows to 3%, a record low

Bloomberg Intelligence said that after DeepSeek released V4.1 Flash, the extent to which China’s top AI models trail their U.S. counterparts in benchmark testing narrowed to 3%。

# Industry Insights # Artificial Intelligence # AI Model # DeepSeek # benchmark testing

MarketNews10/01, 04:07

Google opens access to a limited group of partners Gemini 4 Argon, with mixed internal assessments of coding performance

Google has made Gemini Argon available to a small number of cybersecurity partners Gemini 4 Argon, and plans to offer it first to paying users after testing. Some insiders believe its coding performance does not fully reflect its benchmark results.

# Industry Insights # Artificial Intelligence # Gemini 4 Argon # Coding performance # Benchmarks

ProjectsUnverified09/29, 16:04

The Databricks team said AI the system ranked NVIDIA SOL first in all four tracks of -ExecBench

The Databricks team said its AI system ranked NVIDIA SOL of -ExecBench, 235 kernel tests,GPU first in all four tracks, with model Token costs of approximately 70000 USD。

# Industry Insights # AI programming # GPU Kernel optimization # Large-model inference # Benchmark testing

FrankieInvestment and Research

09/25, 12:09

X account: we gave claude access to a fully autonomous biofoundry to design, execute, and evaluate experiments in synthetic biology. we are also introducing MirrorBench: a new benchmark that measures progress toward the autonomous creation of self-replicating mirror life.

X account @frankieislost posted: “we gave claude access to a fully autonomous biofoundry to design, execute, and evaluate experiments in synthetic biology. we are also introducing MirrorBench: a new benchmark that measures progress toward the autonomous creation of self-replicating mirror life.”

# Synthetic biology # Autonomous experimentation # Biosecurity # Mirror life # Benchmarking

All loaded8 shown · 12 per batch