GPT-5 launched in August 2025 with a 400,000-token context window, multimodal text-and-image input, test-time compute, and access for about 700 million ChatGPT users worldwide. GPT-5 comes in three tiers—GPT-5, GPT-5 Mini, and GPT-5 Nano—sharing the 400K context and multimodal input while offering different performance and cost profiles. The GPT-5 API pricing is $1.25 per million…
Read more
Gemini Ultra achieved 90.0% on the MMLU benchmark, the first AI to exceed human expert performance. GPT-4 Omni (GPT-4o) was introduced by 2025, handling text, images, and audio with about 232 ms response times and powering the free ChatGPT tier in mid-2025. Claude Opus 4 is billed as the world’s best coding model, achieving 72–73%…
Read more
In May 2025, Prime Intellect unveiled INTELLECT-2, a 32B-parameter LLM trained on a globally distributed swarm of volunteer GPUs. INTELLECT-2 is the first model of its scale trained via fully asynchronous reinforcement learning across hundreds of heterogeneous, permissionless machines on chakra.dev. PRIME-RL coordinates the training loop by separating experience generation on inference workers from policy…
Read more
Apple is negotiating with Anthropic and OpenAI to license their large-language models to power a major Siri upgrade, signaling a shift away from Apple’s traditional in-house AI development. The talks include training versions of these models to run on Apple-owned private cloud infrastructure for internal testing. If finalized, the plan would replace Apple’s own Foundation…
Read more
