The program offers training and tools to help entrepreneurs use ChatGPT Work for automation and growth.
What it means for you ↓Why this moves the Board:
The other developments worth knowing, and what each one means for you.
Moonshot's 2.8-trillion-parameter model lands near the top of the field, with open weights that let anyone run it.
Microsoft will tap Mistral's Europe-based data centers to serve cloud customers who want more control.
New fast, cheap models arrived while the flagship Gemini 3.5 Pro slipped again.
The company says its new model trails only Anthropic's top system.
the workhorse tier most apps use gets smarter and cheaper, which can lower your costs.
regulators will be able to fine model makers, so AI in Europe comes with more accountability.
A capable new coding model at $2 and $6 per million tokens, well below Anthropic and OpenAI.
The Chinese lab now charges double during weekday busy hours, a first for its API.
An independent index graded nine frontier labs and found none better than mediocre.
Transient but real, limit resets, brief outages, weekend promos. Verified sources only; the one surface that updates through the day.
The score is only as trustworthy as its receipts. Here's exactly what changed each number this month.
A model's weights are the millions of numbers it learned in training, its actual brain. When a lab releases them as open weights, anyone can download the model and run it on their own computers for free, instead of renting it through an API. Today's Kimi K3 is the largest open-weights model yet. It does not mean the training data or code is public, only the finished model.
Developers are buzzing about GPT-5.6's work agent and a wave of giant open models from China; safety researchers are more sober after a weak industry scorecard.
Reception color from Reddit / HN / X, capped in the Score, never a rank input.
The Vikshy Score in 60 seconds, what it measures, what it ignores, and why a loud launch can't buy the #1 spot.