Tag
This technical report introduces ZAYA1-8B, a mixture-of-experts reasoning model trained on AMD hardware that achieves competitive performance on math and coding benchmarks using under 1B active parameters. It also details Markovian RSA, a novel test-time compute method for aggregating parallel reasoning traces.
User testing shows the new 5.4 Pro behaves like a stealth 5.5-thinking-xhigh, hinting OpenAI is quietly giving Pro subscribers early access to stronger reasoning.
OpenAI introduces GPT-Rosalind, a frontier reasoning model specifically designed to support research in biology, drug discovery, and translational medicine.
OpenAI is releasing GPT-5.4 and GPT-5.4 Pro across ChatGPT, the API, and Codex, featuring native computer-use capabilities, 1M token context, improved reasoning and coding, and state-of-the-art performance on professional knowledge work benchmarks. It is described as OpenAI's most capable and token-efficient reasoning model to date.
OpenAI releases GPT-5.4 Thinking, the latest reasoning model in the GPT-5 series with enhanced safety mitigations, notably the first general-purpose model implementing comprehensive cybersecurity safeguards.
OpenAI introduces GPT-5.2, the most capable model series yet, with significant improvements in knowledge work, code generation, image perception, long-context understanding, and tool-calling. The GPT-5.2 Thinking variant achieves state-of-the-art performance on professional benchmarks, outperforming human experts on 70.9% of GDPval tasks across 44 occupations.
OpenAI releases GPT-5.1, a new model in the GPT-5 series that dynamically adapts thinking time based on task complexity, offering 2-3x faster performance than GPT-5 while maintaining frontier intelligence. The release includes extended prompt caching (24-hour retention), new coding tools (apply_patch and shell), and a 'no reasoning' mode for latency-sensitive applications.
OpenAI releases gpt-oss-120b and gpt-oss-20b, open-weight reasoning models under Apache 2.0 license designed for agentic workflows with strong instruction following, tool use, and chain-of-thought capabilities. The release includes comprehensive safety evaluations confirming the models do not reach high capability thresholds for biological, chemical, or cyber risks even under adversarial fine-tuning.
OpenAI releases o3 and o4-mini, its latest reasoning models that can agentically access and combine all ChatGPT tools (web search, code execution, image analysis, image generation). o3 achieves state-of-the-art performance on coding, math, and science benchmarks with 20% fewer major errors than o1, while o4-mini offers efficient reasoning optimized for cost and speed.
OpenAI releases o3-mini, a cost-efficient reasoning model with strong STEM capabilities, available in ChatGPT and API with support for function calling, structured outputs, and three reasoning effort levels. The model matches o1 performance in math and coding while being faster and cheaper, with free plan users gaining access to a reasoning model for the first time.
OpenAI releases o1 model to API with production-ready features including function calling, structured outputs, vision capabilities, and 60% lower latency than o1-preview. Additional developer tools include Realtime API improvements, Preference Fine-Tuning, and new Go and Java SDKs.
OpenAI released o1, a new series of reasoning-focused AI models that outperform previous models on complex tasks in science, coding, and mathematics. The preview model solved 83% of IMO problems compared to GPT-4o's 13%, and reached the 89th percentile in competitive coding.
OpenAI releases o1-mini, a cost-efficient reasoning model that matches o1 performance on STEM tasks like math and coding while being 80% cheaper. The model is optimized for reasoning-heavy applications and is now available to API users and ChatGPT Plus/Team/Enterprise/Edu subscribers.
OpenAI introduces the o1 model series, designed to reason through complex tasks before responding, with applications in genetics, science, coding, and math. The announcement highlights use cases in decoding genetics with researcher Catherine Brownstein.