Benchmarking voice agents is difficult because it requires evaluating multiple dimensions simultaneously, including latency, response quality, and reasoning ability

Overview Debates Related claims Topics

OpenAI announced that an internal artificial intelligence model refuted the exponential time hypothesis using a proof verified in the Lean proof assistant

Neither OpenAI's nor Anthropic's announced refutation of the exponential time hypothesis had undergone academic peer review at the time of the announcement

Anthropic announced that Claude refuted the exponential time hypothesis on the same day OpenAI made a similar claim

As of October 2026, OpenAI is developing an investment banking evaluation benchmark designed to assess the automation of junior analyst workflows

GPT-6 showed little performance improvement on existing financial benchmarks due to saturated evaluation tasks, ambiguous prompts, and erroneous answer keys

OpenAI discontinued tracking the GDP eval benchmark because the benchmark became saturated and flawed

Software engineering evaluation relies on a combination of verifiable compiler checks and subjective human preferences for maintainable code structure

OpenAI trained generative slide presentation models by utilizing human evaluators to perform pairwise aesthetic comparisons

Artificial intelligence models struggle more with subjective human taste and formatting nuance than with objective factual accuracy

Early reinforcement learning from human feedback relied on pairwise comparison signals to train artificial intelligence models on human preferences

Artificial intelligence models exhibit lower rates of factual hallucination than observed during 2022 and 2023

Legal AI platform Eve built autonomous voice AI agents to place outbound phone calls to insurance companies and retrieve client records

The gap between the top open-weights AI model and the top United States proprietary AI model widened from 3 points to 12 points on Artificial Analysis Intelligence Index Version 5

Artificial Analysis removed long-context reasoning benchmark AA-LCR from the Intelligence Index Version 5

Kimi 3.1 Pro is the top-ranked open-weights model on Artificial Analysis Intelligence Index Version 5 with a score of 46

Google's Gemini 4 Carbon gained 8 points on Terminal Bench 4.0 and 6 points on Frontier Suite compared to Gemini 4 Argon

Claude Haiku 5.5 scored 36 points on the Artificial Analysis Intelligence Index Version 5

Google's Gemini 4 Carbon scored 59 points on the Artificial Analysis Intelligence Index Version 5, ranking in first place ahead of Opus 5.5

Google's Gemini 4 Carbon recorded a 12-point increase on Exploit Gym and a 3-point increase on the CWE Benchmark compared to Gemini 4 Argon

Artificial Analysis added Frontier Code 2.0, a private benchmark evaluating real-world coding pull requests, to the Intelligence Index Version 5