Mistral Releases a Massive Open Model. API Vendors Should Be Nervous.
Mistral released Mixtral 8x22B under an Apache 2.0 license. The model uses a sparse mixture-of-experts architecture. It runs locally on consumer GPUs and matches closed frontier models on coding and reasoning benchmarks.
This teaches the principle of local-first AI deployment: you trade convenience for sovereignty and cost control. Your thinking must shift from 'which API?' to 'can I host this myself?' Small businesses can now eliminate recurring inference fees entirely.
Mistral, a French AI company, released this model. They previously released Mixtral 8x7B. The 8x22B variant competes with larger closed models on coding tasks.
Step 1: Install Ollama from ollama.com, a consumer tool for running local models. Step 2: Open your terminal and run 'ollama run mixtral' to download and launch the smaller 8x7B version, as 8x22B requires substantial GPU memory most consumers lack. Step 3: Ask it a coding question you would normally send to ChatGPT, compare the speed and quality, and note that no API key or internet connection was required after download.