Large Language Models

Summary: Gemini 3.7 Flash in 30 Seconds Gemini 3.7 Flash landed on August 13, 2026, only three weeks after 3.6 Flash. Introductory pricing is $0.75 in / $3.75 out per 1M tokens: half of 3.6 Flash’s standard rate. The promo ends December 31, 2026. Coding jumped hard: FrontierCode 1.1 went from 34.4% to 43.6%, DeepSWE v1.1 from 48.6% to 65.3%. Context window is 1,048,576 tokens with a 65,536 token output cap. It takes text, images, video, audio and PDF. It is not the smartest model on the board: 56 on the Artificial Analysis index, one point behind GPT-5.6 Terra and Muse Spark 1.2, at roughly a third of their price. Google’s Flash line has never been about winning the intelligence crown. It is the model you run at volume: cheap enough to call thousands of times a day, fast enough to sit inside a product. Gemini 3.7 Flash stretches that definition. ...

Summary: DeepSeek V4 Pro in 30 Seconds DeepSeek V4 Pro 0813 went generally available on August 13, 2026. No press release, no blog post, just one line in a changelog. 1.7 trillion parameters, mixture-of-experts architecture, 1M token context, 384K token output ceiling. The weights are on Hugging Face under the MIT license. You can download them, modify them and ship them in a commercial product. Official scores are bold: 87.9 on Terminal Bench 2.1, 62.7 on DeepSWE. Independent testing is more restrained: 53 on the Artificial Analysis Intelligence Index. Pricing is $0.435 in and $0.87 out per million tokens. From August 16 the off-peak rate is half of that. The catch: its smaller sibling V4 Flash scores 52 on the same index at one third of the price. You are paying triple for one point. Model launches have settled into a ritual: a teaser video, a wall of benchmark charts, an excited founder post. DeepSeek skipped all of it. DeepSeek V4 Pro 0813 went generally available on August 13, 2026 with no announcement at all. One line landed in the API changelog and the weights appeared on Hugging Face. ...

Summary: Grok 4.6 in 30 Seconds Grok 4.6 is SpaceXAI’s new flagship model, announced on August 12, 2026. It is a post-training upgrade on the Grok 4.5 base, not a new foundation model. It scores 61 on the Artificial Analysis Intelligence Index: five points above Grok 4.5 from a month ago, level with GPT-5.6 Sol, two points behind Claude Opus 5. Pricing did not move: $2 per 1M input tokens and $6 per 1M output. Roughly half what rivals charge. A new reasoning tier arrived: xhigh. The model also checks and verifies its own work far more often. The context window is 500K tokens, but any request above 200K doubles the price of the whole call. In the AI race, a month is a long time. In early July we were writing about Grok 4.5. Five weeks later SpaceXAI announced Grok 4.6 and put the model straight back into the frontier conversation. ...

Summary: GPT-5.6-Cyber in 30 Seconds GPT-5.6-Cyber is a version of GPT-5.6 Sol trained to be far more permissive on cybersecurity work. Announced on August 10, 2026. It is not publicly available. Only verified security firms and researchers in the Daybreak Red tier can use it. In testing it answered 95% of advanced cyber requests. Standard GPT-5.6 Sol answered just 1.5%. It sits at the High capability level under OpenAI’s Preparedness Framework. The delayed Astra model is the one at the Critical threshold. OpenAI has announced a noticeably less restricted model for cyber defenders: GPT-5.6-Cyber. It is a variant of GPT-5.6 Sol tuned for cybersecurity workflows, and it is closed to ordinary ChatGPT users. ...

Summary: Qwen3.8-Max in 30 Seconds Qwen3.8-Max is Alibaba’s new flagship model, made generally available on August 2, 2026. 2.4 trillion parameters, 95 billion active (MoE architecture). It reads text, images and video, and returns text. Context window is in the 1 million token class. On one task it ran 125 hours (about 5 days) with no human input, rebuilding an experiment from a machine learning paper from scratch, confirming its six findings, then inventing a method that beats the paper. It beats Claude Opus 4.8 on most coding and agent tests, trades blows with Claude Fable 5 and GPT-5.6 Sol, and falls behind on some. API pricing is $2 input / $6 output per million tokens. Repeated input costs $0.25. This is the first time Alibaba has open-weighted a Max-class model. The weights landed on Hugging Face on August 12, 2026, though under Alibaba’s own Qwen3.8-Max license rather than Apache 2.0. Two days later, on August 14, Qwen3.8-27B followed: a dense 27B model under Apache 2.0 that fits on a single GPU. Ask an AI to “rebuild the experiment in this paper, then improve on it” and it normally stalls after a few turns, waiting for you to step in and steer. ...