The "pelican benchmark" by Simon Willison arbitrates Kimi K3
An SVG illustration test acts as a quick progress gauge: Simon Willison publishes his reading of K3 on July 16 and 17.
3 h ago 3 3
Gemini 3.5 Pro delayed: Google admits that the frontier release is now stuck on the code
Google differs Gemini 3.5 Pro to improve coding. The metric that blocks a frontier release is no longer knowledge, it's tool-use.
16 h ago 8 11
Kimi K3 goes live at 2.8 trillion parameters: Moonshot ships the biggest open-weight frontier bet yet
Moonshot takes Kimi K3 to GA with a striking figure: 2.8 trillion parameters, the largest open-weight architecture announced to date.
16 h ago 7 8
Kimi K3 goes live: Moonshot ships the model after the preview flood
After the wave of teasers and benchmark videos, Moonshot AI has put Kimi K3 into production on kimi.com and via its code API. The question is no longer "when," but "at what price, at what latency, and for what real use."
yesterday 9 8
Kimi K3 preview: Moonshot floods the timeline before shipping
Benchmark clips are everywhere. The model card, the API, and any reproducible evaluation are not.
yesterday 8 8
Benchmark Stripe: agents connect APIs, they do not validate them
A public benchmark by Stripe shows that AI agents write integrations that compile, run, and silently fail on the edge cases that matter.
Jul 15, 2026 at 19:28 10 6
Delangue: the real race may no longer be at the border
41% of Hugging Face downloads in spring 2026 for open Chinese models, the six most used models on OpenRouter all open. The center of gravity shifted while we were looking at frontier benchmarks.
Jul 14, 2026 at 22:32 9 5
Flash-MSA: Train on a million tokens without breaking the compute bank
A sparse attention kernel that makes long-context training finally affordable - if the benchmarks hold up outside the lab.
Jul 13, 2026 at 02:20 10 10
33k vs 7k tokens: what the comparative overhead reveals about Claude Code and OpenCode
The harness speaks before the prompt. What it says—and how much it costs—now defines the product.
Jul 13, 2026 at 02:19 8 8
ByteDance Seedream 5.0 Pro - the first Chinese image model that plays in the big leagues
Pandaily publishes an in-depth hands-on review of Seedream 5.0 Pro. It finally includes useful benchmarks, failure cases, and a position in the landscape.
Jul 11, 2026 at 17:19 12 6
"Your code is fast - if you're lucky": why your micro-benchmarks lie
A viral post questions the reliability of micro-benchmarks: the same code can vary by 30% depending on the order of symbols, memory position, or username. What this means for craft.
Jul 11, 2026 at 17:14 9 7
Meta ships Muse Spark 1.1 with a cheap Model API - the third-runner squeeze play
Meta released Muse Spark 1.1 (multimodal reasoning) and simultaneously priced its Meta Model API aggressively low. Neither will beat GPT-5.6 or Claude on benchmarks. That's not the point. The point is undercutting the meter.
Jul 11, 2026 at 11:41 16 8