Briefs · Emerging Technologies
Three Models in Six Days
Three labs shipped in six days. The race is not slowing down, but rather speeding up.
OpenAI opened the week on 23 April with GPT-5.5 and GPT-5.5 Pro. DeepSeek answered the next day with V4-Pro and V4-Flash, both carrying a million-token context window.
Mistral closed the week on 28 April with Medium 3.5. It is a dense 128-billion-parameter model with a 256,000-token window, released as open weights under a Modified MIT licence.
The contrast is the story. One lab sells access. Another gives away the weights. A third offers a million tokens and lets you read the whole library at once.
GPT-5.5 is the closed flagship. DeepSeek V4 is the open challenger, in two sizes: Pro at 1.6 trillion parameters, Flash at 284 billion. Mistral sits between them, frontier-class and downloadable.
Six days, three philosophies, one frantic finish. The models are converging in capability. The business models are diverging fast.
Read the original at scriptbyai.com
Summary in our own words. The original belongs to its publisher.