FrontierQwen released 27 billion under Apache 2.0, with a 2.4 trillion giant next to itOn 3 August we wrote that Alibaba was showing Qwen3.8-Max and promising its weights the following week. The promise has been kept: on 8 August the giant goes up for download, and three days before it, a small 27 billion model under Apache 2.0.FrontierAnthropic touched the guard, not the model - and biological refusals fell by about 85%Since 7 August, Fable 5 answers health and study questions far more often. The change isn't in the model itself, but in the rules of the classifier that decides what counts as dual-use. Virology, toxicology and molecular design stay closed.FrontierFree ChatGPT moves to a stronger model, with unlimited text chatFrom 6 August, paid users get a retuned model with a slider for how deep it thinks. Free users switch to GPT-5.6 Luna by default this week, and get unlimited text conversations from next week. Files and images stay capped.FrontierShieldstral: Mistral releases an open guard model that reads your rules at the moment of the checkMistral released Shieldstral - a European, multimodal safety classifier with 3 billion parameters and open weights under Apache 2.0. The moderation policy goes in as plain text at the call itself, with no retraining. According to Mistral, the model holds its own against systems up to 7 times larger, and does it on a single card with 16GB of memory.FrontierLFM2.5: 2.6 billion parameters trained for agent work right on the deviceLiquid AI released LFM2.5-2.6B - a model built from the ground up for agents on phones and laptops: it follows instructions, handles tools, delivers 220 tokens per second on an Apple M5 Max CPU, and the weights are free on Hugging Face. According to the company, it leads its class on most agentic benchmarks.FrontierAlibaba showed its biggest model - and promised its weights for next weekQwen3.8-Max carries 2.4 trillion parameters and a context of one million tokens, already available through the API. The real news is different: Alibaba says that in days it will release the weights freely - a top-class model anyone will be able to download.FrontierOpenAI's next model is called Astra, and it debuts with ten math resultsOpenAI published ten results on longstanding open problems in mathematics - achieved, by the company's own account, by an internal version of Astra, their next big model. Every proof is formalized in Lean, so a machine can check it. The tokens for all ten cost about $2000.FrontierGrok Imagine learns consistency: one face and one voice in every scenexAI added references to the Imagine Video 1.5 model: you send a photo of a character and a voice recording, and the model keeps them the same from scene to scene. Text-only video and native 1080p resolution are also coming. References launch first in the US, for the paid tiers.
80 stories