Posted on 09/02/2026, 07:41 By zgzsur
| OPEN AI NEW CODEX |
| 1. OpenAI launches GPT-5.3 Codex and introduces the OpenAI Frontier platform. OpenAI presents GPT-5.3 Codex, its most powerful code model yet, achieving new records in SWE-Bench Pro (57%) and Terminal-Bench 2.0 (77%), while also being 25% faster and saving more tokens than its predecessors. Simultaneously, OpenAI announces Frontier, a platform designed for enterprises to design, deploy, and pilot AI agents capable of performing real-world operational tasks. The agents share a common context, benefit from structured onboarding, and have clear permissions to increase efficiency. |
| 2. Anthropic presents Claude Opus 4.6, its flagship model for enterprise use. Anthropic launches Claude Opus 4.6, its most advanced model to date, designed for intellectual work within the enterprise. It achieves top scores in programming, financial analysis, and legal reasoning—skills now integrated into Claude Cowork . The model better plans its responses, requires less guidance in complex workflows, and offers a context window of 1 million tokens for extended sessions. Regarding Claude Code, Anthropic also introduces “agent teams”: a preview that allows multiple AI agents to collaborate on a common goal—planning, executing, and delivering—without the need for step-by-step instructions. |
| 3. Perplexity launches Model Council to compare responses from top AI models. Perplexity introduces Model Council, a feature that runs the same query on three state-of-the-art AI models in parallel (such as Claude Opus 4.5, GPT-5.2, and Gemini 3 Pro). A synthesis model then analyzes the responses, arbitrates discrepancies, and produces a single, consolidated, and more reliable answer. The goal: to help users validate and compare the quality of responses from the most advanced models on the market. |
Categories: Android