Anthropic says Claude Sonnet 5.5 is faster while keeping token prices unchanged
Anthropic says Sonnet 5.5 keeps Sonnet 5 pricing while cutting latency and task cost, but its docs also warn developers about five breaking changes before upgrading.
AI Models
The capabilities, limitations and choices behind today's AI models.
Latest stories
22 stories
Google launches Gemini 4 Argon with limited access for trusted cyber defenders
Google says Gemini 4 Argon is rolling out first to trusted cyber defenders through Fairwind while broader access waits for safety testing. The model is aimed at long-horizon coding and cybersecurity, but its benchmark claims are company-reported.
OpenAI launches GPT-6.1 Sol at one-fifth GPT-6 Astra’s token price
OpenAI has released GPT-6.1 Sol, a lower-cost model that the company says nears GPT-6 Astra on coding and agentic work. The published pricing tables show standard text tokens at one-fifth of Astra’s rates, with important deployment limits.
AMD’s $8.2 billion World Labs deal ties 3D world models to its AI roadmap
AMD said it will acquire World Labs for about $8.2 billion, bringing Fei-Fei Li’s spatial-intelligence team into its AI roadmap. The deal underscores growing interest in 3D world models, robotics simulation and model-chip co-design.
PrismML shows a 1-bit model running on Snapdragon smart glasses
PrismML says its 1-bit Bonsai model ran locally on Snapdragon AR1 Gen 1 smart glasses at Snapdragon Summit, with claimed gains in memory use and speed. The demo is promising, but no shipping smart glasses running PrismML have been announced.
OpenAI and Anthropic Launch Cheaper AI Models: What Buyers Can Verify
OpenAI and Anthropic introduced lower-priced models, but the verifiable story is narrower: published API rates, model access and supported surfaces are confirmed; performance claims still come from the vendors.
How to choose among OpenAI’s model tiers by cost, capability and task
OpenAI’s model docs now present a clear ladder: GPT-6 Astra for hardest reasoning, GPT-5.6 Sol/Terra/Luna for cost-controlled general work, and specialized models for narrow tasks like cyber, speech and vision.
Before you download an open model, read its model card
Downloadable weights are only part of the decision. A model card helps you check intended use, evidence and what the release leaves unexplained.
Choose an AI model with a task, not a leaderboard
A useful model comparison starts with your own work. Here is a practical way to judge accuracy, effort and cost without mistaking a demo for evidence.