2026-07-30

2026-07-30에 감지된 AI 모델 가격·컨텍스트·종료일 변경 전체입니다.

변경모델내용
종료 예정z-ai/glm-4.5v2026-12-31 종료 · D-150 남음
가격 변경moonshotai/kimi-k2.6출력 +47.1% · 캐시 읽기 +47.1% · 입력 +47.1%
가격 변경openai/gpt-5.6-luna출력 -80.0% · 캐시 읽기 -80.0% · 캐시 쓰기 -80.0% · overrides · 입력 -80.0%
가격 변경openai/gpt-5.6-luna-pro출력 -80.0% · 캐시 읽기 -80.0% · 캐시 쓰기 -80.0% · overrides · 입력 -80.0%
가격 변경openai/gpt-5.6-terra출력 -20.0% · 캐시 읽기 -20.0% · 캐시 쓰기 -20.0% · overrides · 입력 -20.0%
가격 변경openai/gpt-5.6-terra-pro출력 -20.0% · 캐시 읽기 -20.0% · 캐시 쓰기 -20.0% · overrides · 입력 -20.0%
가격 변경z-ai/glm-5.2출력 +57.4% · 캐시 읽기 +57.4% · 입력 +57.4%
가격 변경~moonshotai/kimi-latest출력 -6.7%
가격 변경deepseek/deepseek-chat출력 +28.6% · 입력 +28.6%
가격 변경google/gemma-4-31b-it출력 -15.0% · 캐시 읽기 과금 시작 · 입력 -28.6%
가격 변경nvidia/nemotron-3-ultra-550b-a55b출력 +63.6% · 캐시 읽기 +100.0% · 입력 +20.0%
가격 변경qwen/qwen-2.5-7b-instruct출력 +100.0% · 입력 +150.0%
가격 변경qwen/qwen3-vl-30b-a3b-instruct출력 +15.4% · 입력 +15.4%
가격 변경z-ai/glm-5.2출력 -11.4% · 캐시 읽기 -11.4% · 입력 -11.4%
가격 변경~moonshotai/kimi-latest캐시 읽기 -3.3% · 입력 -3.3%
모델 삭제openai/gpt-5-codex
모델 삭제openai/o3-deep-research
모델 삭제openai/o4-mini-deep-research
당사자 발언Sam Altman @sama우리는 모든 수준에서 최고의 가격/지능 절충점을 제공하고 싶다 || we want to offer the best price/intelligence tradeoff at every level
당사자 발언Sam Altman @sama오늘 대대적인 가격 인하: *80% 인하: GPT-5.6 Luna, 이제 million input tokens당 $0.20, million output당 $1.20 *20% 인하: GPT-5.6 Terra, $2/$12로 *GPT-5.6 Sol은 API에서 Fast mode를 제공하며, 최대 2.5x의 속도를 2x의 가격에, 동일한 지능으로 제공합니다 https://t.co/erC6u4VoDR || major price cuts today: *80% drop for GPT-5.6 Luna, now $0.20 per million input tokens and $1.20 per million output *20% drop for GPT-5.6 Terra, to $2/$12 *GPT-5.6 Sol gets Fast mode in the API, up to
당사자 발언OpenAI @OpenAI고급 지능을 더 풍부하고 저렴하게 만드는 일은 AGI가 모든 인류에게 혜택을 주도록 보장하려는 우리의 사명에서 핵심입니다. GPT-5.6 Sol의 도움으로 우리는 효율성 면에서 큰 도약을 이루었습니다. 오늘 우리는 Luna와 Terra의 가격 인하를 통해 그 성과를 API에 반영하고 있으며, https://t.co/TieMXHtxSd || Making advanced intelligence more abundant and affordable is central to our mission to ensure AGI benefits all of humanity. With the help of GPT-5.6 Sol, we have made leaps in efficiency. Today, we ar
당사자 발언OpenAI @OpenAI또한 ChatGPT app과 Codex CLI의 Auto-review를 GPT-5.4에서 GPT-5.6 Luna로 업그레이드합니다. Luna의 새로운 가격과 함께, Auto-review 비용이 약 10x 낮아져 에이전트형 워크플로를 더 비용 효율적으로 만들 것으로 예상합니다. || We’re also upgrading Auto-review in the ChatGPT app and Codex CLI from GPT-5.4 to GPT-5.6 Luna. Combined with Luna’s new price, we expect Auto-review to cost about 10x less, making your agentic workfl
당사자 발언OpenAI @OpenAIGPT-5.6 Luna와 Terra의 가격 인하와 함께, API의 GPT-5.6 Sol용 Fast mode는 Standard processing보다 최대 2.5x 빠른 속도를 Standard price의 2x에 제공합니다. Fast mode는 지능의 변화 없이 API 고객이 GPT-5.6 Sol에 더 빠르게 접근할 수 있게 해줍니다. || Along with the price reduction on GPT-5.6 Luna and Terra, Fast mode for GPT-5.6 Sol in the API delivers up to 2.5x the speed of Standard processing at 2x the Standard price. Fast mode gives API custom
당사자 발언OpenAI @OpenAI우리는 비용 효율성, 역량, 속도 전반에서 모델의 최전선을 넓히는 데 전념하고 있습니다. 오늘부터 GPT-5.6 Luna의 가격을 80%, GPT-5.6 Terra의 가격을 20% 인하하고, API에서 GPT-5.6 Sol을 위한 더 빠른 옵션을 제공합니다. Luna와 Terra의 낮아진 가격은 https://t.co/rFhK7XKedp || We are committed to pushing the model frontier across cost efficiency, capability, and speed. Starting today, we are reducing prices for GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20% , and offering a f
당사자 발언Google AI Developers @googleaidevsGemini API 또는 @GoogleAIStudio를 사용해 지금 더 똑똑하고 안전한 물리적 에이전트를 구축하세요 ↓ https://t.co/30bgUlQlSP || Build smarter, safer physical agents today using the Gemini API or @GoogleAIStudio ↓ https://t.co/30bgUlQlSP
당사자 발언Google AI Developers @googleaidevsGemini Robotics ER 2는 물리 AI를 위해 설계된, 당사의 가장 뛰어난 체화 추론 모델입니다 🤖 로보틱스를 위한 고수준 두뇌로 구축된 이 모델은 Gemini Live API에 직접 연결됩니다. 연속 비디오 스트림을 처리해 진행 상황을 추적하고, 도구를 호출하고, 웹을 검색하며, https://t.co/ukAufrrggY || Gemini Robotics ER 2 is our most capable embodied reasoning model designed for physical AI 🤖 Built as a high-level brain for robotics, the model connects directly to the Gemini Live API. It processes
OUTAGEDegraded performance on Claude Opus 4.8
ANNOUNCEMENTGemini Robotics 2 brings whole body intelligence to robots
ANNOUNCEMENTGemini Robotics ER 2: powering robotics with video understanding, task orchestration, and multi-robot collaboration
RELEASElangchain-core==1.5.3
RELEASEb10197
RELEASEv2.51.0
RELEASEv2.16.0
ANNOUNCEMENTLimit remote control to managed devices
ANNOUNCEMENTStacked pull requests are now in public preview
ANNOUNCEMENTGitHub Copilot in Visual Studio — July update
ANNOUNCEMENTReference same-repository actions with self-repository syntax
ANNOUNCEMENTStacked sessions and pull requests in the GitHub Copilot app
ANNOUNCEMENTGPU Management: Why Idle GPUs Are the New Grounded Aircraft
ANNOUNCEMENTAdvancing the price-performance frontier with GPT-5.6
OUTAGEElevated errors affecting ChatGPT conversations
파라미터 변경openai/gpt-5.1-codex-max삭제 max_tokens
파라미터 변경openai/gpt-5.2-codex삭제 max_tokens
파라미터 변경thinkingmachines/inkling추가 response_format
당사자 발언Sam Altman @sama이거 정말 기대된다. 과학적 발견을 크게 가속할 모델에 아주 가까워졌다. 이를 위한 가장 좋은 방법은 우리가 모든 것을 스스로 알아내려 하는 게 아니라 과학자들에게 힘을 실어주는 것이다. 우리 모두는 그 혜택을 누릴 자격이 있다. https://t.co/nzfHpb96re || so excited for this. very close to models that will significantly accelerate scientific discovery; the best way to do this is for us to empower scientists, not to try to figure out everything ourselve
당사자 발언OpenAI @OpenAI이러한 실험이 평가에서는 모델만을 단독으로 측정하는 경우가 드물며, API 설정, 하네스 설계, 프롬프팅에 관한 눈에 덜 띄는 여러 선택도 함께 측정한다는 사실을 상기시키는 계기가 되기를 바랍니다. 성능을 극대화하려는 API 개발자라면, 다음을 사용하는 것을 권장합니다 || We hope these experiments serve as a reminder that evals rarely measure models in isolation—they also measure a bundle of less visible choices about API settings, harness design, and prompting. If you
당사자 발언OpenAI @OpenAI벤치마크 점수는 모델뿐 아니라 모델을 실행하는 데 사용된 하네스와 설정도 반영합니다. 장시간 실행되는 에이전트의 경우, 추론을 유지하고 컨텍스트를 압축하면 모델이 이미 학습한 내용을 바탕으로 작업을 이어갈 수 있습니다. https://t.co/JBIKCKrUfw || A benchmark score reflects the model as well as the harness and settings used to run it. For long-running agents, retaining reasoning and compacting context lets the model build on what it has already
당사자 발언OpenAI @OpenAIResponses API로 하네스를 구현하고 다음 기능을 켰습니다. → 추론 유지 → 컨텍스트 압축 공개 세트에서 GPT-5.6 Sol의 점수는 출력 토큰을 6x 더 적게 사용하면서 188% 상승했습니다. https://t.co/uN1IrKEugu || We implemented the harness with the Responses API and turned on: → Retained reasoning → Context compaction On the public set, GPT-5.6 Sol’s score rose 188% while using 6x fewer output tokens. https://
당사자 발언OpenAI @OpenAIARC-AGI-3는 모델이 지시 없이 낯선 2D 게임을 얼마나 잘 학습할 수 있는지 테스트합니다. 표준 하네스는 각 동작 후 GPT-5.6 Sol의 추론을 버렸고, 컨텍스트가 차오르면서 이전 행동도 제거했습니다. 모델은 계속 처음부터 다시 시작해야 했습니다. https://t.co/cLtHYMNqiO || ARC-AGI-3 tests how well models can learn unfamiliar 2D games without instructions. The standard harness discarded GPT-5.6 Sol’s reasoning after each move and dropped earlier actions as the context fi
당사자 발언OpenAI @OpenAIGPT-5.6 Sol은 수학의 미해결 문제를 푸는 데 사용되어 왔습니다. 그렇다면 2D 퍼즐 게임 벤치마크인 ARC-AGI-3에서는 왜 고전하고 있었을까요? 저희가 조사했습니다. 하네스가 모델이 학습한 내용을 기억하지 못하게 하고 있었습니다. 두 가지 API 설정을 활성화하자 점수가 세 배로 올랐다는 것을 발견했습니다. https://t.co/0mv05MnSn2 || GPT-5.6 Sol has been used to solve open problems in mathematics. So why was it struggling with ARC-AGI-3, a benchmark of 2D puzzle games? We investigated. The harness was not letting it remember what
OUTAGEElevated errors across many models
OUTAGEElevated errors across all models
RELEASEb10189
ANNOUNCEMENTCopilot code review: Agent skills and MCP now generally available
OUTAGECopilot model Claude Fable 5 experiencing elevated errors
OUTAGEIncident with Copilot AI Model Providers
ANNOUNCEMENTHow GPT-5.6 fuses frontier intelligence with frontier efficiency
ANNOUNCEMENTHow enabling two settings tripled our scores on the ARC-AGI-3 benchmark

이전 2026-07-29 · 다음 2026-07-31