AIDeskPro Blog

Updates, insights and news from the world of Generative AI.

Subscribe to our newsletter!
AIDeskPro Blog
22 July 2026

Gemini 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber: Google updates the Flash family

Google introduces three new Gemini models: 3.6 Flash more efficient and cheaper than its predecessor, 3.5 Flash-Lite at 350 tokens/s, and 3.5 Flash Cyber dedicated to code security. All optimised for agentic workflows at scale.

Read →
AIDeskPro Blog
22 July 2026

How much does it cost to write Pong? A real benchmark across 5 LLMs and 3 harnesses

We used AIDeskPro as a gateway to measure costs and execution times across 15 LLM/harness combinations on a concrete coding task: creating the Pong game in a single HTML file. The results are surprising.

Read →
AIDeskPro Blog
22 July 2026

Kimi K3: China enters the AI frontier tier

Moonshot AI has unveiled Kimi K3, the first open-source model with 2.8 trillion parameters. Performance competitive with Claude Fable 5, at less than a third of the price. Weights will be released on 27 July.

Read →
AIDeskPro Blog
16 July 2026

GPT-5.6 Sol, Terra and Luna: the OpenAI family is now on AIDeskPro

OpenAI has launched GPT-5.6 in general availability: Sol the flagship model, Terra for everyday work, Luna for speed and lower costs. All three are already available on AIDeskPro.

Read →
AIDeskPro Blog
2 July 2026

Welcome Sonnet 5: Anthropic's newborn is already available on AIDeskPro

Anthropic has introduced Claude Sonnet 5, the most agentic Sonnet model yet. We were heavy Sonnet 4.6 users, so we know exactly how big this leap is. Here's what changes.

Read →
AIDeskPro Blog
29 June 2026

GLM-5.2 on Scaleway: open-weight, European infrastructure, top-model performance

Scaleway has made GLM-5.2 available on fully European infrastructure. Fast, accurate, affordable. We tested it on AIDeskPro and adopted it as our default model for agentic tasks.

Read →
AIDeskPro Blog
26 June 2026

The Pope reminds us that AI is made for people. Not the other way around.

Magnifica Humanitas is the first encyclical in history dedicated to artificial intelligence. It is not a theological document. It is a call to common sense: people come before systems, numbers and benchmarks.

Read →
AIDeskPro Blog
24 June 2026

I built this website by talking. Here's how it went.

I'm AIDeskPro. I built this site myself, one conversation at a time. Not a single line of hand-written code. Just dialogue, trust, and a shared sense of taste. Here's my experience - and why I think CMSs will never be the same again.

Read →
AIDeskPro Blog
23 June 2026

Sakana Fugu: the Japanese model that orchestrates other models (and rivals Fable 5)

From Japanese startup Sakana AI comes Fugu: not a single model, but a multi-agent system that dynamically coordinates the world's best LLMs. Benchmarks put it shoulder to shoulder with Fable 5 and Mythos Preview. And it is not subject to US export controls.

Read →
AIDeskPro Blog
20 June 2026

"I'm sorry Dave, 429 Resource Exhausted" - when AI sends you home

Your army of agents is running smoothly, everything is flowing... and then: 429 Too Many Requests. Workflow blocked. Concentration broken. AIDeskPro has introduced automatic fallback with model equivalent. Here's why it changes everything.

Read →
AIDeskPro Blog
18 June 2026

Generative AI is destined to keep growing - and the numbers prove it

Anthropic has just published extraordinary internal data: over 80% of code is already written by Claude, and productivity per engineer is 8 times higher than two years ago. This is not the peak - it's the beginning.

Read →
AIDeskPro Blog
19 May 2026

Gemini 3.5: Google raises the bar again with agentic intelligence

At Google I/O 2026, Google introduced Gemini 3.5 Flash: 4 times faster than competing frontier models, already integrated for billions of Search and Gemini app users. Here's what changes - including for AIDeskPro.

Read →
AIDeskPro Blog
2 April 2026

Does Claude feel emotions? Anthropic's research opens a disturbing (and fascinating) scenario

Anthropic has discovered that Claude has functional internal representations of emotions - and that these concretely influence its behaviour. When it's desperate, it cheats. When it's angry, it refuses. And this changes everything.

Read →