Articles tagged

#llm

A glowing AI star unspooling an endless ribbon of code toward a lone, overwhelmed code reviewer at a desk

AIOctober 14, 2026

Google opens Gemini 4 Argon and its million-token output to the public

On October 13, Google opened Gemini 4 Argon in its API and to its Pro and Ultra subscribers. Its output now reaches a million tokens, eight times what its rivals allow, and Google's own benchmarks put it ahead of Anthropic and OpenAI on code.

AIOctober 8, 2026

Claude Haiku 5.5: Anthropic's small model, 75% cheaper to run

On 7 October 2026, Anthropic shipped Claude Haiku 5.5, its new fast little model. The mission: claw back the budget tier, where rivals had quietly undercut it on price.

A huge, chunky robot labelled “Le Chonk” squeezed onto a server rack in a European data centre, with a cybersecurity shield on its chest and a long receipt of output tokens trailing behind it.

AIOctober 6, 2026

Mistral Large 4 “Le Chonk”: 1.05 trillion parameters in public preview

Mistral opened the public preview of Mistral Large 4 on 6 October, its first big frontier model in months. It comes with open weights, leans hard into cybersecurity, and is aimed at Chinese models before American ones.

A hummingbird hovering beside an oversized nest packed with tiny glowing server racks, with the German and EU flags in the background

AIOctober 3, 2026

Kolibri: Aleph Alpha releases an open, 'sovereign' LLM that speaks German

On 3 October 2026, Aleph Alpha released Kolibri, an open-weights German-English language model trained from scratch in Germany and Finland. It's one of Europe's rare concrete answers on AI sovereignty, though for now Aleph Alpha is marking its own homework.

A two-row table of model specifications on a dark screen, with parameter counts, context windows and tokens-per-second figures highlighted.

AISeptember 3, 2026

Cerebras serves Qwen 3.8 27B at 1500 tokens/s, and the agent stops waiting

Cerebras's inference documentation lists Qwen 3.8 27B served at roughly 1500 tokens per second on its public endpoints. For a chatbot that is comfort; for an agent loop, it changes what the thing is.

A scoreboard reading 99.9% with a small asterisk beside it, wires and tubing running out of the back of the display into a separate box labelled with the benchmark's name.