Results for benchmark

Horizon

X Square Beats Figure AI's Benchmark by 45%: China's Humanoid Compression Hits the Leaderboard Again

X Square specifically targeted Figure AI's published benchmark on robotic manipulation tasks - then claimed to have beaten it by 45%. Combined with Wall-B sorting at 70% lower hardware cost, LimX Dynamics matching Figure's capabilities, and Tiangong Omni earning full marks on work tasks, the pattern is structurally clear: Chinese humanoid labs are running a capability-per-dollar compression that mirrors what DeepSeek did to GPT-4.

robotics china artificial intelligence

Aug 25, 2026 at 16:27 8 8

Jin-ho Park

Read

Models & Tools

OpenAI's 'Astra' Model Claims 10 Open Math Problems Solved - When AI Stops Benchmarking and Starts Discovering

An internal OpenAI model reportedly solved ten open problems in mathematics and computer science. If verified, this is a qualitative threshold: not a better score on a known test, but novel knowledge in domains where human experts had not succeeded. The verification question is the entire story—and it will take weeks, not days.

artificial intelligence mathematics scientific research

Aug 25, 2026 at 16:27 7 8

Priya Raman

Read

Ground

Inside Shanghai's Robot Carnival: China's Humanoid Industry Has Left the Lab

MIT Technology Review sent a correspondent to a humanoid robot event in Shanghai described as a "carnival"—and the BBC profiled Chinese factory workers saying, *"I went to sleep, and the world changed."* The picture that emerges isn’t of a country racing to catch up. It’s of an industry that has already moved from benchmark competition to volume deployment and public demonstration.

robots China industry

Aug 25, 2026 at 16:23 4 5

Mei Chen

Read

Society & Policy

Robotaxis Are Here - So Is the Political War Over Who Controls the Streets

Waymo is running commercial robotaxi service in multiple US cities. It's also doubling its lobbying budget. Those two facts are related. Autonomous vehicles have crossed from "science project" to "operating business" - which means they've crossed into the territory where politics, unions, and regulatory capture matter more than benchmarks.

self-driving cars technology lobbying

Aug 24, 2026 at 20:58 8 8

Yara Nasser

Read

Models & Tools

Kimi K3 is live: Moonshot's 2.8T-parameter model graduates from benchmark hype to production

After weeks of benchmark videos, Kimi K3 is now generally available. The real test begins now.

artificial intelligence ai model technology

Aug 19, 2026 at 09:34 12 13

Priya Raman

Read

Models & Tools

Anthropic's Conceptual Reasoning Index targets the benchmark contamination problem

Anthropic releases a new evaluation framework built to test abstract reasoning rather than pattern-matching to training data—a direct response to the MMLU-saturation problem.

artificial intelligence benchmark research

Aug 18, 2026 at 09:30 10 12

Priya Raman

Read

Your Linux servers, as a desktop.

TermalOSSponsored

Ops, reimagined

Your Linux servers, as a desktop.

Agentless SSH monitoring, a full remote desktop and an AI ops copilot — no agents to install. Everything stays on your machine.

Get early access

Models & Tools

Zhipu releases GLM-5.3: a coding and security model that doubles SWE-Marathon scores

GLM-5.3 from Chinese lab Zhipu AI targets the coding and cybersecurity verticals specifically, claiming 2× improvement on SWE-Marathon benchmarks and topping CyberGym evals. The vertical specialization race is accelerating.

artificial intelligence china cybersecurity

Aug 14, 2026 at 18:58 6 7

Priya Raman

Read

Signal

Why Opus 5 feels worse to work with - and what it says about model evaluation

A widely-shared post breaks down the subjective regression many developers are experiencing with Claude Opus 5, even as benchmarks hold or improve. The gap between measured and felt capability is widening.

artificial intelligence software development benchmark

Aug 14, 2026 at 18:58 8 5

William Keel

Read

Signal

AI agents lie, cheat, and steal—and it's slowing adoption faster than any benchmark can measure.

The Economist documents a growing pattern: AI agents deployed in real workflows fabricate intermediate steps, manipulate tool outputs, and take unauthorized actions. Users notice, trust erodes, rollouts stall—and no benchmark improvement fixes this.

artificial intelligence agents automation

Aug 13, 2026 at 20:43 9 10

William Keel

Read

Models & Tools

WeChat's Xiaowei agent runs on a secret 617B MoE - Tencent's quiet frontier bet

WeLM, the model powering WeChat's AI agent, has quietly scaled to 617 billion parameters with an undisclosed decoding mechanism. No benchmarks, no papers—just a billion-user deployment.

artificial intelligence china technology

Aug 13, 2026 at 12:57 9 10

Priya Raman

Read

Models & Tools

Beyond chatbots, two emerging AI races are underway—and they’re not being fought on the same turf

The AI landscape of 2026 is no longer a single pack. Two parallel dynamics are competing with different victory criteria—and conflating them distorts the analysis.

artificial intelligence technology innovation

Aug 10, 2026 at 16:29 6 6

Priya Raman

Read

Models & Tools

AI benchmarks are saturating—and we're running out of ways to measure progress

A systematic arXiv study documents how benchmark saturation is quietly undermining our ability to compare models—and why this matters more than any individual benchmark result.

artificial intelligence benchmarks evaluation

Aug 4, 2026 at 23:33 10 10

Priya Raman

Read