2026-09-04에 감지된 AI 모델 가격·컨텍스트·종료일 변경 전체입니다.
| 변경 | 모델 | 내용 |
|---|---|---|
| 종료일 공지 | z-ai/glm-4.7-flash | - -> 2026-09-10 |
| 가격 변경 | deepseek/deepseek-v4-flash | 출력 +8.5% · 캐시 읽기 +8.5% · 입력 +8.5% |
| 가격 변경 | deepseek/deepseek-v4-pro | 출력 -3.1% · 캐시 읽기 -3.1% · 입력 -3.1% |
| 가격 변경 | deepseek/deepseek-v4-pro-0813 | 출력 -12.2% · 캐시 읽기 -12.2% · 입력 -12.2% |
| 가격 변경 | nvidia/nemotron-3-ultra-550b-a55b | 출력 +30.2% · 캐시 읽기 +56.2% · 입력 +4.2% |
| 가격 변경 | qwen/qwen3.5-35b-a3b | 출력 -40.0% · 캐시 읽기 과금 중단 · 입력 -68.0% |
| 가격 변경 | qwen/qwen3.6-27b | 출력 -44.4% · 캐시 읽기 -75.0% · 입력 -50.0% |
| 가격 변경 | undi95/remm-slerp-l2-13b | 입력 -22.2% |
| 가격 변경 | ~deepseek/deepseek-v4-flash-latest | 출력 -37.5% · 캐시 읽기 -23.1% · 입력 -0.0% |
| 가격 변경 | ~z-ai/glm-latest | 출력 -3.0% · 캐시 읽기 -47.0% · 입력 +0.2% |
| 신규 모델 | inclusionai/ling-3.0-flash-sante:free | 컨텍스트 262,144 |
| 신규 모델 | nvidia/nemotron-3.5-content-safety | 컨텍스트 131,072 |
| 신규 모델 | openai/gpt-6-astra | 컨텍스트 1,050,000 |
| 신규 모델 | openai/gpt-6-astra-pro | 컨텍스트 1,050,000 |
| 신규 모델 | x-ai/grok-4.3:batch | 컨텍스트 1,000,000 |
| 모델 삭제 | ibm-granite/granite-4.1-8b | |
| 당사자 발언 | Tibo @thsottiaux | @jrysana 테스트해 봤는데 Plus 구독으로 충분히 할 수 있어요. Astra에는 10-15 minutes가 걸리고 주간 사용량의 15% 미만을 사용합니다. 이는 주당 ~6개의 테이블을 판매할 수 있다는 뜻입니다. 또는 한 달에 24개의 테이블입니다. 테이블당 85 cents 미만이 됩니다. || @jrysana I tested it and you can comfortably do it in the Plus subscription. It takes Astra 10-15 minutes and less than the 15% of the weekly usage. This means that you can sell ~6 tables per week. Or |
| 당사자 발언 | OpenAI @OpenAI | GPT-6 Astra는 오늘 제한된 수의 조직을 대상으로 출시되며, 앞으로 며칠에 걸쳐 모든 ChatGPT Plus, Pro, Business, Enterprise 사용자가 이용할 수 있게 됩니다. 또한 OpenAI API와 AWS를 통해서도 이용할 수 있습니다. 최상의 Astra를 경험할 준비를 하세요. ChatGPT를 받으세요 https://t.co/6u0irS8iF0 || GPT-6 Astra is rolling out today to a limited set of organizations and over the coming days will become available to all ChatGPT Plus, Pro, Business, and Enterprise users, as well as through the OpenA |
| 당사자 발언 | OpenAI @OpenAI | GPT-6 Astra는 FrontierMath Tier 4, ARC-AGI 3, TerminalBench-4.0에서 최첨단 수준입니다. GPT‑6 Astra는 과학적 발견에서도 중대한 진전을 이루었으며, Terminal-Bench Science 0.1과 HealthBench Pro에서 최첨단 성능을 보입니다. https://t.co/7hEFadVAN9 || GPT-6 Astra is state-of-the-art on FrontierMath Tier 4, ARC-AGI 3, and TerminalBench-4.0. GPT‑6 Astra is also a major advance for scientific discovery, with state-of-the-art performance on Terminal-Be |
| 당사자 발언 | OpenAI @OpenAI | Astra는 사용자 의도를 이해하는 능력이 크게 향상된, 가장 정렬이 잘 된 모델입니다. https://t.co/rY0An7o3rq || Astra is our most aligned model, with substantial improvements in understanding user intent. https://t.co/rY0An7o3rq |
| 당사자 발언 | OpenAI @OpenAI | Astra는 여러 전문 직종의 컴퓨터 워크플로 작업을 위한 벤치마크인 Agents’ Last Exam, AutomationBench, ScreenSpot Pro에서 최첨단 결과를 달성합니다. https://t.co/DiWogwT6Dx || Astra achieves state-of-the-art results on Agents’ Last Exam, AutomationBench, and ScreenSpot Pro, benchmarks for computer workflow tasks across professions. https://t.co/DiWogwT6Dx |
| 당사자 발언 | Anthropic @AnthropicAI | 주요 수학적 증명이 올바른지 확인하는 데 수년이 걸릴 수 있습니다. 형식화—수학적 추론을 Lean과 같은 컴퓨터 증명 보조기가 검증할 수 있는 형태로 변환하는 것—가 도움이 될 수 있습니다. 지난달 Claude는 그중 하나인 Fermat’s Last Theorem의 최초 형식화된 증명을 완료했습니다 https://t.co/pdT8zwlV4A || Checking that a major mathematical proof is correct can take years. Formalization—converting the mathematical reasoning into a form computer proof assistants like Lean can verify—can help. Last month, |
| RELEASE | v2.1.260 | |
| RELEASE | 0.153.3 | |
| RELEASE | v3.8.0 | |
| ANNOUNCEMENT | GitHub Actions: Early September 2026 updates | |
| ANNOUNCEMENT | Gemini 3.8 Flash is now available in GitHub Copilot | |
| ANNOUNCEMENT | GitHub CLI Linux package signing key expires September 5 | |
| ANNOUNCEMENT | Multiple trusted publishing configurations for npm | |
| ANNOUNCEMENT | Upcoming deprecation of selected GitHub Copilot models | |
| ANNOUNCEMENT | Project HydraFusion: Frontier quality via multi-model orchestration | |
| ANNOUNCEMENT | Daybreak for Frontline Defenders: $1B to protect essential services | |
| ANNOUNCEMENT | GPT-4 | |
| ANNOUNCEMENT | Legora reviewed 41 documents in minutes with GPT-6 Astra | |
| ANNOUNCEMENT | Playco cut manual fixes 50% prototyping games with GPT-6 Astra | |
| ANNOUNCEMENT | Safety overview: GPT-6 Astra | |
| OUTAGE | Users in APAC region may face increased error in ChatGPT, Work, image generation, file upload, Voice, and Codex Cloud | |
| 메타데이터 변경 | inclusionai/ling-3.0-flash | |
| 메타데이터 변경 | inclusionai/ling-3.0-flash-fin | |
| 메타데이터 변경 | inclusionai/ling-3.0-flash-fin:free | |
| 파라미터 변경 | meta-llama/llama-3.1-70b-instruct | 삭제 logprobs, top_logprobs |
| 파라미터 변경 | upstage/solar-pro-3 | 추가 parallel_tool_calls |
| 파라미터 변경 | upstage/solar-pro4 | 추가 parallel_tool_calls |