Governor Gavin Newsom has vetoed SB 1047, which would have imposed safety requirements on the largest AI developers. He said its model-size threshold did not adequately measure real-world risk.
OpenAI introduces o1-preview and o1-mini, two models designed to spend more time on internal reasoning. The company is targeting math, science, and programming, although the launch comes with usage limits and missing features.
OpenAI is adding Structured Outputs to its API so GPT-4o and GPT-4o mini can respond according to a JSON schema defined by the developer. The feature reduces a common problem when connecting models to applications and databases.
Google has rolled out an improved version of Gemini 1.5 Pro with better math and code performance, and cut API pricing by up to half to attract more developers.
Mistral AI unveils Large 2, a 123-billion-parameter model focused on code, math and multilingual work. Its weights are released for research, and the model is available through the company’s API and major cloud platforms.
Meta has released Llama 3.1 in 8-billion, 70-billion and 405-billion-parameter versions. The largest model brings top-tier capabilities within developers’ reach, although its license still imposes usage conditions.
Anthropic introduces Claude 3.5 Sonnet, a model that outperforms Claude 3 Opus on several tests despite costing far less. It also debuts Artifacts, a workspace for working with generated code and documents.
DeepSeek has released Coder V2, an open family of programming and mathematics models based on a mixture-of-experts architecture. Its largest version has 236 billion parameters, though it activates 21 billion per query.
Mistral AI has launched Codestral, a 22-billion-parameter model focused on generating and completing code. It enters a market dominated by coding assistants, but its licensing terms restrict how downloadable weights can be used.
Snowflake has introduced Arctic, a mixture-of-experts language model with 480 billion total parameters and an Apache 2.0 license. It is designed to generate SQL, code and useful answers in enterprise environments.
Startup Cognition has unveiled Devin, an AI agent that takes on coding tasks, creates a plan and works with a terminal, editor and browser. The company says it outperforms other systems on the SWE-bench benchmark.
Google DeepMind introduces Gemini 1.5 Pro, a model that handled up to one million context tokens in tests. The leap lets it analyze documents, video, audio and code at an unusual scale.
This website uses cookies to improve the browsing experience. Cookie policy.
No results. Try different words.
