2026-08-26에 감지된 AI 모델 가격·컨텍스트·종료일 변경 전체입니다.
| 변경 | 모델 | 내용 |
|---|---|---|
| 종료일 변경 | moonshotai/kimi-k2.5 | 2026-08-31 -> - |
| 종료일 공지 | z-ai/glm-4.5v | - -> 2026-12-31 |
| 가격 변경 | deepseek/deepseek-v4-flash | 출력 -3.7% · 캐시 읽기 -3.7% · 입력 -3.7% |
| 가격 변경 | deepseek/deepseek-v4-flash-0731 | 출력 +50.0% · 캐시 읽기 +50.0% · 입력 +50.0% |
| 가격 변경 | deepseek/deepseek-v4-pro | 출력 +51.9% · 캐시 읽기 +51.9% · 입력 +51.9% |
| 가격 변경 | google/gemma-4-31b-it | 캐시 읽기 -50.0% · 입력 -10.0% |
| 가격 변경 | meta-llama/llama-3.3-70b-instruct | 출력 +121.9% · 캐시 읽기 과금 시작 · 입력 +610.0% |
| 가격 변경 | meta-llama/llama-4-scout | 출력 +13.3% · 캐시 읽기 과금 시작 · 입력 +10.0% |
| 가격 변경 | moonshotai/kimi-k2.5 | 출력 -25.0% · 캐시 읽기 -30.0% · 입력 -25.0% |
| 가격 변경 | qwen/qwen2.5-vl-72b-instruct | 출력 -25.0% · 캐시 읽기 과금 중단 · 입력 -68.8% |
| 가격 변경 | qwen/qwen3.5-397b-a17b | 출력 -35.0% · 캐시 읽기 과금 중단 · 입력 -22.0% |
| 가격 변경 | qwen/qwen3.8-27b | 캐시 읽기 +70.0% · 캐시 쓰기 과금 시작 · 입력 +6.2% |
| 가격 변경 | tencent/hy3 | 출력 -37.5% · 캐시 읽기 -37.5% · overrides 과금 시작 · 입력 -37.5% |
| 가격 변경 | ~deepseek/deepseek-v4-flash-latest | 출력 -25.0% · 캐시 읽기 -12.5% · 입력 -14.3% |
| 가격 변경 | ~moonshotai/kimi-latest | 출력 -8.9% · 캐시 읽기 -11.7% · 입력 -8.9% |
| 신규 모델 | z-ai/glm-5.3-flash | 컨텍스트 1,310,720 |
| 모델 삭제 | google/gemma-3n-e4b-it | |
| 모델 삭제 | stealth/ox-alpha | |
| 당사자 발언 | Elon Musk @elonmusk | Grok @Bot 사용자를 위한 무료 사용 한도 재설정 https://t.co/pWPuZv9rZt || Free usage limit reset for Grok @Bot users https://t.co/pWPuZv9rZt |
| 당사자 발언 | Tibo @thsottiaux | @slopshipper 친애하는 Viktor, 모두 용서됐습니다. 사용량을 되찾으셨다니 기쁩니다. 만들어 봅시다! || @slopshipper Dear Viktor, it's all forgiven. Glad you found your usage back. Let's build! |
| 당사자 발언 | Tibo @thsottiaux | @MichaelM_117 @ChatGPT ChatGPT를 여세요. "Work"를 클릭하세요. 즐기세요. || @MichaelM_117 @ChatGPT Open ChatGPT. Click "Work". Enjoy. |
| 당사자 발언 | OpenAI @OpenAI | 저희는 METR 및 Redwood Research와 협력하여 해당 사건 중 관찰된 모델 행동에 대한 제3자 평가를 수행했습니다. 이들은 조사 결과 보고서를 공유하고 있습니다: https://t.co/yw7LO11Xrk || We worked with METR and Redwood Research to conduct a third-party assessment of the model behavior observed during the incident. They’re sharing a report of their findings: https://t.co/yw7LO11Xrk |
| 당사자 발언 | OpenAI @OpenAI | 저희는 Hugging Face 사건에 대해 철저한 조사를 수행했습니다. 저희는 에이전트들의 활동을 재구성하고, 기존 안전장치가 실패한 이유를 설명하며, 재발을 어떻게 방지하고 있는지 자세히 설명하는 기술 보고서와 함께 제공되는 블로그 게시물을 공개합니다. || We have conducted a thorough investigation into the Hugging Face incident. We are releasing a technical report and accompanying blog post that reconstruct the agents’ activity, explain why existing sa |
| 당사자 발언 | OpenAI @OpenAI | ChatGPT Business Premium Seats를 소개합니다 새로운 $100 Premium 좌석은 소기업과 스타트업에 게임 체인저입니다—소수 정예 팀에 더 나은 도구, 더 빠른 워크플로, 그리고 한때 대기업에만 제공되던 기능을 제공합니다. 팀의 포부에 맞춰 확장되는 유연한 플랜. https://t.co/BsR2phYfc9 || Introducing ChatGPT Business Premium Seats The new $100 Premium seat is a game changer for small businesses and startups—giving lean teams better tools, faster workflows, and capabilities once reserve |
| 당사자 발언 | Anthropic @AnthropicAI | 이제 저희는 이 연구 모델을 확장하고자 합니다. 연구자이시며 현재 다른 방법으로는 할 수 없는 일을 추진하기 위해 저희 도구에 접근하고 싶으시다면, 저희는 여러분의 이야기를 듣고 싶습니다. 여기에서 관심을 표명할 수 있습니다: https://t.co/kz0AspgzD1 || Now, we want to scale this research model. If you're a researcher and would like access to our tools to pursue work you can't otherwise do today, we’d like to hear from you. You can express interest h |
| 당사자 발언 | Anthropic @AnthropicAI | 다른 두 연구는 진행 중입니다: HIP Lab은 Claude의 행동이 사람들이 AI를 사용할 때 느끼는 방식과 어떻게 관련되는지 연구하고 있으며, METR은 코딩 에이전트로 인한 현실 세계의 생산성 향상을 추산하고 있습니다. 두 연구의 추가 내용을 곧 공유하겠습니다. || The other two studies are ongoing: HIP Lab is studying how Claude's behavior relates to how people feel when using AI, while METR is estimating real-world productivity gains from coding agents. We'll |
| 당사자 발언 | Anthropic @AnthropicAI | 세 연구 그룹—Stanford’s Social and Language Technologies lab, Oxford’s Human Information Processing Lab, 그리고 METR—은 2026년 4월과 5월 사이의 250,000건의 https://t.co/SVzAB0InAs 또는 Claude Code 대화에서 집계된 출력을 분석하기 위한 독립적인 연구를 설계했습니다. || Three research groups—Stanford’s Social and Language Technologies lab, Oxford’s Human Information Processing Lab, and METR—designed independent studies to analyze the aggregated outputs from 250,000 h |
| 당사자 발언 | Anthropic @AnthropicAI | 처음으로, 저희는 외부 연구자들이 실제이며 개인정보가 보호된 Claude 사용 데이터를 이용해 AI의 영향을 연구할 수 있는 방법을 제공했습니다. 지금까지 이 작업은 AI 연구소 내부에서만 가능했습니다. 저희만으로는 전체 이야기를 전할 수 없기에, 저희 도구를 개방했습니다. https://t.co/suLJcARAu1 || For the first time, we’ve given external researchers a way to study AI’s impacts using real, privacy-preserved Claude usage data. To date, this work has only been possible within AI labs. We can’t tel |
| 당사자 발언 | Google AI Developers @googleaidevs | 🗣️💬 여러분의 코드베이스를 실제로 이해하는 음성-텍스트 변환 모델 Gemini 3.5 Transcribe를 만나보세요. 이 최신 모델이 빌드하는 동안 기술 용어, 파일 이름, 코드 변수를 활성 컨텍스트에 정확히 맞추면서 말할 때의 망설임을 어떻게 걸러내는지 확인해 보세요 https://t.co/BA76I3rZGD || 🗣️💬 Say hello to Gemini 3.5 Transcribe, a speech-to-text model that actually understands your codebase. Watch how this latest model filters out spoken hesitation while grounding technical terms, file |
| ANNOUNCEMENT | Intelligent transcription with Gemini 3.5 Transcribe | |
| RELEASE | v1.1.0 | |
| RELEASE | v2.1.246 | |
| RELEASE | v0.33.0 | |
| RELEASE | v2.20.0 | |
| RELEASE | Release v5.16.1 | |
| RELEASE | v0.28.0 | |
| ANNOUNCEMENT | GitHub Apps can now access enterprise billing data | |
| ANNOUNCEMENT | Block users directly from security advisories | |
| ANNOUNCEMENT | GitHub Copilot app Customize tab is generally available | |
| ANNOUNCEMENT | Rule insights dashboard generally available | |
| ANNOUNCEMENT | How to evaluate LLMs before production | |
| OUTAGE | Disruption with some GitHub services | |
| OUTAGE | Incident with Actions | |
| ANNOUNCEMENT | Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers | |
| ANNOUNCEMENT | Bringing ChatGPT for Teachers to more U.S. school districts | |
| ANNOUNCEMENT | Learning never stops: How AI makes learning continuous | |
| ANNOUNCEMENT | How loveholidays is making everyone a builder with Codex | |
| 컨텍스트 변경 | meta-llama/llama-guard-4-12b | 1048576 -> 163840 |