List of AI News about benchmarks
| Time | Details |
|---|---|
|
2026-11-19 12:52 UTC |
Google DeepMind SVP says no single AGI benchmark, focus on reliability
According to Google DeepMind via its X post featuring SVP Koray Kavukcuoglu and researcher Shane Legg, AGI progress will be judged by steadier, real‑life model reliability rather than one definitive benchmark. |
|
2026-11-04 15:05 UTC |
AI Dev NYC features harness engineering workshops
According to DeepLearning.AI, Hugging Face engineer Joel Niklaus will lead two AI Dev NYC workshops on harness engineering and benchmarks, with speakers from OpenAI, Neo4j, Oracle, and Sonar on Nov 30–Dec 1 in New York. |
|
2026-11-01 20:34 UTC |
SemiAnalysis launches GDPx agent benchmark; top score 12%
According to SemiAnalysis via X, the new GDPx benchmark measures agentic capabilities across high‑value domains, where even the best model scores only 12%, highlighting significant headroom and cost considerations noted by industry leaders. |
|
2026-10-28 09:17 UTC |
Google unveils Gemini 4 Carbon, tops AI index in latest analysis
According to The Rundown AI citing Artificial Analysis, Google’s Gemini 4 Carbon scores 59 on the Intelligence Index, edging Claude Opus 5.5 and leading 14 of 16 benchmarks, with early access limited to select cybersecurity defenders. |
|
2026-10-28 09:03 UTC |
Gemini 4 Carbon claims SOTA on key AI benchmarks: Analysis
According to Demis Hassabis on X, Google DeepMind's Gemini 4 Carbon achieves state of the art across key benchmarks and integrates frontier capabilities with deep safety and security research. |
|
2026-10-26 02:30 UTC |
Canva unveils first in-house image model climbing benchmarks
According to The Rundown AI, Canva launched its first in-house image generation model, which is climbing industry benchmarks and ranking lists for image quality and performance. |
|
2026-10-08 16:00 UTC |
AREX Harness Boosts Benchmarks by 10 Points
According to DeepLearningAI, AREX verifies requirements, scores 82.5 BrowseComp and 82.0 F1 WideSearch en, and its harness alone adds up to 10 points. |
|
2026-10-07 10:30 UTC |
Claude Haiku 5.5 Breakthrough beats Luna
According to @TheRundownAI, Anthropic’s Claude Haiku 5.5 is faster, cheaper, and tops Luna on benchmarks, starting at $0.10 per million input tokens. |
|
2026-09-25 14:36 UTC |
OpenAI AGI Claims Spark 2020 Flashback
According to @timnitGebru, OpenAI researchers said GPT3 achieved AGI in 2020, raising scrutiny of benchmarks and model evaluation standards. |
|
2026-09-24 09:49 UTC |
METR Chart Replacement Sparks 2026 Analysis
According to @emollick, a new visual should replace METR’s Long Task Horizon chart, prompting fresh analysis of AI capability benchmarks. |
|
2026-09-24 08:47 UTC |
Claude Sonnet 4.5 gains with Meta memory agents
According to DeepLearningAI, Meta pairs action and memory agents to curb context rot, boosting Claude Sonnet 4.5 accuracy from 37.6% to 45.9%. |
|
2026-09-16 06:40 UTC |
AI benchmarks Crisis Exposed by New Paper
According to @emollick, public AI benchmarks are saturated or error ridden, severely underestimating current model abilities, as the cited paper shows. |
|
2026-09-15 06:30 UTC |
Claude Fable 5.1 tops v4.2 Index
According to DeepLearning.AI, Claude Fable 5.1 leads Intelligence Index v4.2 with 57, ties OpenAI’s new model on benchmarks, and cuts agent costs via cache. |
|
2026-09-10 05:29 UTC |
DeepSeek Causal Encoder‑Decoder Breakthrough
According to KyeGomezB, DeepSeek’s causal encoder‑decoder MoE cuts active params to 8B in, 16B out, promising lower cost and top benchmarks. |
|
2026-08-27 11:24 UTC |
Terminal-Bench-Science Launches benchmark analysis
According to StanfordAILab, Terminal-Bench-Science v0.1 tests 70 research tasks and Claude Opus 5 solves about 30%, signaling big gaps in agent tooling. |
|
2026-08-18 19:17 UTC |
Qwen 27B Sparks Debate in Agentic Benchmarks
According to Ethan Mollick, Qwen 27B is a game changer, but critics on X urge benchmarking, noting weaker agentic task performance. |
|
2026-08-06 14:58 UTC |
GPT5.6 Signals Personal Finance Breakthrough
According to gdb, OpenAI’s GPT 5.x shows major gains on internal personal finance benchmarks, pointing to stronger planning and advisory performance. |
|
2026-07-24 20:41 UTC |
BenchBenchBench Reveals AI Benchmarking Meta-Analysis
According to emollick, a Codex-prompted project produced BBBBB, an executable benchmark testing AI-authored conformance suites for metrics. |
|
2026-07-24 09:47 UTC |
Claude Opus 5 Breakthrough beats rivals at half cost
According to TheRundownAI, Anthropic’s Claude Opus 5 leaps past 4.8 and outperforms Fable on several benchmarks at half the price. |
|
2026-07-17 01:18 UTC |
Moonshot Kimi K3 Tops Benchmarks, Trails Leaders
According to @CNBC, Moonshot AI says Kimi K3 beats top OpenAI and Anthropic models on select benchmarks while still trailing overall leaders. |