101 articles in "パフォーマンス" — sorted by importance and recency
Claude API experiencing degraded performance for US Midwest traffic
Anthropic's Claude API experienced degraded performance on September 10 affecting users whose traffic routes through the US Midwest. Users reported elevated latency and request timeouts. The incident
Degraded performance detected for Claude Opus 5 and Claude Haiku 4.5
Anthropic reported elevated error rates on requests to Claude Opus 5 and Claude Haiku 4.5 on their status page. As of Aug 19, 09:42 UTC, the root cause was identified and a fix is in progress. No spec
Degraded performance affecting multiple Claude models under investigation
Anthropic reported on Aug 18 at 16:20 UTC elevated error rates and degraded performance affecting Claude Mythos 5, Claude Fable 5, Claude Opus 5, Claude Sonnet 5, Claude Haiku 4.5, and other Claude mo
Degraded performance across multiple Claude models under investigation
Anthropic reported elevated error rates on requests to multiple Claude models as of Aug 12, 13:50 UTC. The company is investigating the issue and promises an update soon. Root cause, scope, and ETA fo
Claude Opus 5 experiencing degraded performance; Anthropic identifies root cause
Anthropic reported elevated error rates on requests to Claude Opus 5 as of August 5. The company has identified the root cause and is actively working on a fix. Updates will be provided as soon as res
Degraded performance on Claude Opus 4.8 — Anthropic investigating
Anthropic reported degraded performance on Claude Opus 4.8 as of July 30 and is actively investigating. The official status page shows the incident is under investigation, though specific impact scope
Elevated errors on Sonnet 4.6 — incident resolved
Anthropic experienced elevated error rates on Sonnet 4.6 on April 24. The incident was reported at 02:55 UTC, identified at 03:09 UTC, and a fix was implemented by 03:15 UTC. Full resolution was confi
Introducing Gemini 3.8 Live with Live Avatar
Google DeepMind announced Live Avatar support for Gemini 3.8 Live, enabling AI to interact via dynamic avatar representations in real time. This evolution from text and static interfaces aims to deliv
OpenAI Introduces GPT-6 Sol and Luna Models
OpenAI has introduced GPT-6 Sol and Luna, two frontier intelligence models designed to serve everyday work with different balances of capability and cost. The announcement marks a strategy to offer ch
Perplexity deploys GPT-6 Astra for end-to-end production systems
Perplexity uses OpenAI's GPT-6 Astra model to automate end-to-end production tasks including writing communications, making software changes, and monitoring systems with significantly reduced human ov
DeepSeek Launches V4.1-Flash: 8x Faster Inference, Lower Costs, Phasing Out V4-Pro
DeepSeek unveiled V4.1-Flash, a 552B-parameter MoE model featuring a new Causal Encoder–Decoder architecture with only 8B active parameters for input and 16B for output. The model reduces KV cache sto
GPT-6 Astra: OpenAI's most capable model for business with advanced reasoning and computer use
OpenAI announced GPT-6 Astra, its most capable model for business. Features advanced reasoning, computer use (automated screen interaction), and stronger writing and design judgment. Targets enterpris
GPT-6 Astra: New generation with state-of-the-art computer use, coding, and security
OpenAI has announced GPT-6 Astra, positioning it as the company's most intelligent and aligned model to date, with state-of-the-art capabilities in computer use, coding, cybersecurity, and science. Sp
Introducing agentic video understanding with Gemini
Google DeepMind introduces agentic video understanding capabilities to Gemini. The new feature enables Gemini to analyze video content more deeply and handle complex tasks involving temporal reasoning
OpenAI Built Realtime Voice AI System 'GPT-Live' in Six Months
OpenAI developed 'GPT-Live,' enabling continuous voice interaction with AI using a turnless speech model and low-latency architecture for faster, more natural conversations. Completed in six months, t
Degraded performance on Claude Sonnet 5 under investigation
Anthropic reported degraded performance on Claude Sonnet 5 on July 31 and is currently investigating the issue. While specific causes and affected features are not detailed in the status update, users
Two API Settings Triple GPT-5.6 Performance on ARC-AGI-3 Benchmark
OpenAI announced that two API configuration adjustments tripled GPT-5.6's scores on the ARC-AGI-3 benchmark by retaining reasoning processes and enabling a new compaction feature. This boosts both per
GPT-5.6 Fuses Frontier Intelligence with Frontier Efficiency
OpenAI unveiled GPT-5.6, improving AI efficiency across models, inference, and agentic workflows. The update aims to deliver more useful intelligence per dollar, balancing performance gains with cost-
Claude Opus 5が新たにリリースされ、コーディングや科学研究の性能が大幅に向上。コストは従来のOpus 4.8と同じで、長期的なエージェント支援に設計されている。これにより、ユーザーからの評価も高まっている。
Introducing Claude Opus 5
Claude Opus 5 has been released, achieving significant performance improvements in coding and professional work. This model is designed to support long-running agents at the same cost as Opus 4.8. It
Luma Runs Production Inference on AMD and Tensorwave
Luma is collaborating with AMD and TensorWave to build multimodal AGI. They utilize AMD's MI325X GPUs for inference on models like Ray3.2, leveraging significant computational resources. The migration
GPT-5.6: Frontier intelligence that scales with your ambition
OpenAI has announced the new model GPT-5.6. This model enhances intelligence per token, strengthens performance per dollar, and provides capabilities to support users in their most challenging tasks.
Introducing Claude Sonnet 5
Claude Sonnet 5 has been launched, showcasing top-tier intelligence for coding and everyday tasks. This most agentic Sonnet model improves significantly over predecessors in reasoning, tool use, and c
FLUX.2 [klein]とFLUX VTOの違いを比較します。新たに発表されたFLUX.2は、高速な画像生成が可能で、消費者向けハードウェアでの利用が実現しました。一方、FLUX VTOはバーチャル試着に特化した機能を持ちます。両者の特徴を見ていきましょう。
FLUX.2 [klein]: Towards Interactive Visual Intelligence
The FLUX.2 [klein] model has been released, offering fast image generation and editing in a single architecture. It runs on consumer hardware with as little as 13GB VRAM, enabling real-time image gene
Sakana AI Releases Fugu Ultra to Compete with Leading Models
Sakana AI's newly released Fugu Ultra matches leading models like Anthropic's Fable 5 and Mythos Preview, delivering frontier capabilities without export control risks. It achieves this through autono
Introducing Gemini Omni
Google DeepMind has announced a new AI model, Gemini Omni. This model features versatility to handle various tasks, especially integrating natural language processing and image recognition. Designed f
Runway Gen-4は、動画生成における新たな一歩を踏み出しました。過去のツールと比較すると、一貫性や効率性が向上しており、クリエイターにとって強力なサポートとなるでしょう。
Runway Unveils Act-One: A New Tool for Character Animation
Runway has announced a new tool, Act-One, designed to create expressive character animations using video and voice performances as inputs. This tool simplifies traditional facial animation workflows,
Runway Gen-4: AI Video Generation with World Consistency
Runway Gen-4 introduces AI-powered video generation technology. This new model aims to enhance consistency in generated visuals, allowing for more realistic and natural videos. Users can create comple
Databricks Integrates GPT-5.5 into Enterprise Agent Workflows
Databricks has adopted GPT-5.5 for enterprise agent workflows. The model set a new state of the art on the OfficeQA Pro benchmark, allowing businesses to streamline their operations more efficiently.
Enhancing FLUX Performance in Collaboration with NVIDIA
Anthropic's new collaboration with NVIDIA significantly enhances FLUX model performance, making it more accessible for creators. This partnership reduces memory requirements, supports a variety of GPU
FLUX.1 Kontext [dev]とFLUX.1 Kontext [pro]を比較しました。新モデルは12Bパラメータで無料提供され、消費者向けに最適化されています。一方、従来のプロモデルはAzure AI Foundryでの利用が可能で、企業向けの高速編集が特徴です。
FLUX.2は、FLUX.1の機能を進化させ、高品質な画像生成を実現しました。特に、複数の参照画像に対するスタイル一貫性や高解像度の画像編集能力が強化されています。開発者向けのオープンウェイトモデルも提供し、クリエイティブワークフローを一新します。
Ray3とターミナルバイオレッタマッチング（TVM）の比較を通じて、動画生成モデルの進化に注目します。Ray3はリアリズムと創造的忠実度を兼ね備えたプロ向けモデルで、TVMはスピードと効率性を重視しています。それぞれの特徴と利点を探ります。
AnthropicがNVIDIAと提携し、FLUXモデルの性能向上を実現。新機能として3D環境向けのサポートが追加され、特にFLUX.1 [dev]は10GBのVRAMで2倍の速度を達成。これに対し、過去のFLUXモデルはAzureでの利用が進んでおり、編集機能が最大8倍速に達しています。
Anthropicの新リリースFLUX.1ツールは、特にインペインティング機能を強化し、FLUXのエコシステムを向上させます。一方、過去のFLUXモデルはAzureでの利用が注目されています。新機能の追加と古いモデルの廃止がどのように影響するかを比較しました。
新たに設立されたBlack Forest LabsがFLUX.1を発表しました。過去のFLUXモデルとの比較を通じて、性能の向上や新機能を明らかにします。特に視覚品質の向上と多様なスタイルに注目です。
FLUX.1 Toolsの新機能が従来のFLUXモデルとどのように異なるのかを比較しました。FLUX.1 Fillの編集機能や廃止された機能について詳しく解説します。特に新しいツールによる性能向上は注目です。
FLUX1.1 [pro]の新バージョンが生成速度を6倍に向上させ、BFL APIが一般提供されることで、開発者は新しい画像生成技術を利用しやすくなります。過去バージョンとの比較を通じて、進化した機能や性能を紹介します。
FLUX.1 Kontextは新たに発表された画像生成モデルで、従来のFLUXモデルと比較して、テキストと画像を融合させたプロンプト生成が特徴です。特に、画像の編集機能が強化され、生成速度も8倍向上しています。これにより、ユーザーはより迅速かつ高品質な画像生成が可能になりました。
Black Forest Labsの設立により、最先端の生成的深層学習モデルFLUX.1が発表されました。FLUXモデルはAzure AI Foundryで利用可能になり、編集機能やリアリズム向上が図られています。本記事では、新旧のFLUXモデルの機能や資金調達の違いを比較します。
FLUX Models Launch on Azure AI Foundry for Enterprise-Ready Image Generation
Black Forest Labs' FLUX models are now available on Microsoft Azure AI Foundry. FLUX.1 Kontext [pro] and FLUX1.1 [pro] are advanced models for text-to-image and image-to-image generation. Notably, FLU
FLUX.2 Launch: Revolutionizing Creative Workflows
Anthropic has announced FLUX.2, which generates high-quality images with consistent style across multiple references. It can read and write complex text, adhere to brand guidelines, and edit images up
Announcement of the new model FLUX.1 Krea [dev]
BFL (Black Forest Labs) announced a new text-to-image model, FLUX.1 Krea [dev], developed in collaboration with Krea AI. This model aims to overcome the typical 'AI look' to achieve photorealism. FLUX
FLUX.1 Kontext [dev] Launches Open Weights for Image Editing
Until now, capable generative image editing models were proprietary tools. The release of FLUX.1 Kontext [dev] changes this, offering a 12B parameter model for consumer hardware. This open-weight mode
Introducing FLUX.1 Kontext and the BFL Playground
FLUX.1 Kontext is a new suite of generative flow matching models for image generation and editing. Unlike traditional text-to-image models, it allows prompts using both text and images, enabling the e
Enhancing FLUX Model Performance through NVIDIA Collaboration
Anthropic has partnered with NVIDIA to enhance the performance of its FLUX models, making them more accessible to a wider community of creators. This collaboration reduces memory requirements, boosts
Release of FLUX.1 Tools for Enhanced Image Control
Anthropic has announced FLUX.1 Tools, a suite designed to enhance control and steerability for its base text-to-image model FLUX.1. FLUX.1 Fill introduces advanced inpainting capabilities that outperf
New High-Resolution Features Added to FLUX1.1 [pro]
FLUX1.1 [pro] introduces new high-resolution capabilities, allowing image generation up to 4MP at just 10 seconds per sample, priced competitively at $0.06. The new 'Ultra Mode' generates images at fo
Announcing FLUX1.1 [pro] and the BFL API
FLUX1.1 [pro] has been released with a 6x faster generation speed compared to its predecessor. The BFL API is now generally available, allowing developers and businesses to integrate cutting-edge imag
Launch of Black Forest Labs and Introduction of FLUX.1 Models
Black Forest Labs has been launched, introducing the state-of-the-art FLUX.1 generative AI models. FLUX.1 excels in text-to-image synthesis, boasting high visual quality and style diversity. The found
Announcing the Launch of Black Forest Labs
Today, we are excited to announce the launch of Black Forest Labs. Our mission is to develop state-of-the-art generative deep learning models for media like images and videos, pushing the boundaries o
FLUX.2 - Next Generation Image Generation
FLUX.2 is the next generation image generation technology developed by Black Forest Labs. It achieves state-of-the-art quality, speed, and controllability in AI image generation. Users can specify par
Introducing FLUX.1 Tools
FLUX.1 Tools has been released, enhancing control over the text-to-image model FLUX.1. The suite includes four features enabling image modification and recreation. Notably, FLUX.1 Fill allows seamless
Introducing FLUX1.1 [pro] Ultra and Raw Modes for Enhanced Image Generation
FLUX1.1 [pro] adds new high-resolution capabilities, supporting up to 4MP images with a generation time of only 10 seconds per sample. The Ultra mode boasts a generation speed over 2.5x faster than co
Ray3 Evaluation Report – State-of-the-Art Performance for Pro Video Generation
Ray3 is a pro-level video generation model that integrates realism, control, and creative fidelity to realize creative intent. Its unique evaluation framework measures performance across multiple dime
Pushing the Limit of Efficient Inference-Time Scaling with Terminal Velocity Matching
Terminal Velocity Matching (TVM) is a new single-stage training paradigm for efficient generation. It achieves the same sample quality while providing a 25x speedup over standard diffusion models. TVM
Ray3 Modify Enhances Video Editing with Luma AI Dream Machine
Luma AI Dream Machine has introduced Ray3 Modify, enabling video editing through natural language. Users can easily remove or swap objects, refine characters, and create virtual sets. With new keyfram
SAM 3.1: Faster and More Accessible Real-Time Video Detection and Tracking
Meta AI has announced the new Segment Anything Model (SAM) 3.1. This model enhances real-time video detection and tracking by increasing speed and accessibility for a broader range of users. The newly
Introducing the Uni-1.1 API: Intelligence You Can Direct
Luma Labs has announced the Uni-1.1 API, a REST interface for image generation and natural language editing. This API enables developers to transform creative workflows and enhance creativity. Uni-1 p
Mistral AIの新モデルMistral 7Bと過去のモデルMistral Large 2を比較。7Bは7.3Bパラメータで推論に強く、一方でLarge 2は多言語対応で128kのコンテキストウィンドウを持つ。各モデルの特長を見ていこう。
Mistral Medium 3.5とMistral Large 2を比較。新モデルはリモートコーディングエージェントや新機能を提供し、タスク処理能力が向上。どちらが優れているのかを探ります。
Stable Diffusion 3.5 NIMのリリースは、企業向けの画像生成を大幅に簡素化し、性能を向上させました。過去の音声生成技術やメディア制作の提携と比較すると、特に企業ニーズに応じたカスタマイズ性が強化されています。
Stability AIが発表したStable Video 4D 2.0は、4D生成技術において大きな進歩を遂げました。特に、単一動画からの品質向上や動的アセットの生成が容易になり、商業利用にも対応しています。過去の技術と比較し、どのように進化したのかを見ていきます。
Mistral AIの新モデル「Mistral Large 2」と前モデル「Mistral Large」を比較します。新モデルはパラメータ数やコンテキストウィンドウのサイズを大幅に向上させ、性能も改善されています。特に多言語対応とコード生成能力が強化され、研究用ライセンスも提供される点が注目されます。
Mistral AIが新たに発表した「Ministral」シリーズは、エッジコンピューティング向けのモデルで、128kのコンテキスト長をサポートします。一方、「Mistral Large」は32Kトークンのコンテキストウィンドウを持ち、多言語タスクに強いです。両者は異なる用途に特化しており、その違いを明確に理解することが重要です。
Mistral AIの新モデル「Mistral Large」と過去のリリースを比較します。Mistral Largeは32Kトークンに対応し、高い推論能力を持つ一方、過去のインターフェースやカスタマイズ機能は異なる方向性を示しています。これにより、ユーザーはより幅広い選択肢を得ることができます。
Stability AIのStable DiffusionモデルがAMD Radeon GPU向けに最適化され、画像生成速度が最大3.8倍向上しました。過去のNVIDIA最適化版と比較すると、AMD版は効率的な動作を実現し、クリエイティブなアプリケーションにおいてさらなる進化を遂げています。
Mistral AIが発表した新しいコードモデル「Codestral」と、過去の言語モデル「Mistral Large」を比較します。両者は異なる用途に特化しており、特にCodestralは多様なプログラミング言語に対応。コンテキストウィンドウのサイズは共通ですが、機能やターゲットユーザーに違いがあります。
ElevenLabsの新モデルTurbo v2と過去のMultilingual V2を比較。Turbo v2はレイテンシと生成速度の向上が特徴で、音質は同等。多言語対応の進展も期待される。
Stable Diffusion 3.5は速度とメモリ効率を大幅に改善し、商用利用にも対応。Stable Virtual Cameraは2Dから3D動画生成を可能にする新技術。両者の特性を比較し、利用シーンを明らかにします。
Googleの新しい第8世代TPUは、特化チップを搭載しAIの処理能力を向上させています。一方、Gemini 3.1 Flash TTSは自然な音声生成を実現。両者は異なる分野での進化を代表しており、AIの未来を形作る重要な技術です。
最新の画像生成モデル「Nano Banana 2」と過去のAI関連技術を比較しました。新モデルは高速生成や一貫性を強化しており、実用性が向上しています。これに対し、過去のモデルは医療や分散トレーニングなど異なる分野での成長を目指していました。
新しいCodestral 25.01は、初代Codestralに比べて約2倍速くコードを生成できるよう改善されました。80以上のプログラミング言語をサポートし、企業向けのローカルデプロイ機能も搭載されています。生産性が大幅に向上するポイントが魅力です。
Introducing Voice Changer: A New Tool for Voice Conversion
ElevenLabs has announced a new voice conversion tool called 'Voice Changer.' This tool allows users to convert a recording of one voice to sound as if spoken by another, preserving the original emotio
ElevenLabs' highly anticipated Turbo v2 model has arrived
ElevenLabs has announced its new speech generation model, Turbo v2, which operates at approximately 400ms latency and is over twice as fast as the previous V1 models. The audio quality matches that of
ElevenLabs Introduces Eleven Multilingual v1 Supporting Seven New Languages
ElevenLabs has launched a new speech synthesis model, Eleven Multilingual v1, which supports seven new languages: French, German, Hindi, Italian, Polish, Portuguese, and Spanish. This advanced model l
Stability AI and Arm Bring On-Device Generative Audio to Smartphones
Stability AI partners with Arm to enable on-device generative audio on smartphones. This allows for high-quality sound effects and audio samples to be created without an internet connection. The techn
Stable Diffusion Now Optimized for AMD Radeon™ GPUs and Ryzen™ AI APUs
Stability AI has collaborated with AMD to release ONNX-optimized versions of the Stable Diffusion models. These models run faster and more efficiently on AMD Radeon™ GPUs and Ryzen™ AI APUs. Available
Stability AI and Arm Launch Stable Audio Open Small for On-Device Audio Generation
Stability AI, in partnership with Arm, has released Stable Audio Open Small as open-source. This 341 million parameter text-to-audio model is optimized to run on Arm CPUs, allowing smartphones to gene
Stability AI Unveils Stable Video 4D 2.0 with Major Enhancements
Stability AI has launched Stable Video 4D 2.0, significantly improving 4D generation and novel view synthesis from a single video. This model offers higher quality outputs for both commercial and non-
Stable Diffusion 3.5 Models Optimized for 2X Speed and 40% Less Memory
Stability AI has optimized Stable Diffusion 3.5 (SD3.5) models with NVIDIA's TensorRT, achieving up to 2.3X faster image generation and a 40% reduction in VRAM requirements. This makes enterprise-grad
Stability AI and NVIDIA Launch Faster Performance with Stable Diffusion 3.5 NIM
Stability AI announces its collaboration with NVIDIA to launch the Stable Diffusion 3.5 NIM microservice. This new service is designed to help enterprises deploy image generation models quickly and ea
Mistral AI Announces New Model 'Mistral Large'
Mistral AI has announced its latest language model, 'Mistral Large', which features top-tier reasoning capabilities and is available through Azure. This model supports complex multilingual tasks and a
Mistral AI Unveils New Model 'Mistral Large 2'
Mistral AI has announced its new model, 'Mistral Large 2.' This model significantly enhances capabilities in code generation, mathematics, and reasoning compared to its predecessor, with improved mult
Mistral AI Introduces State-of-the-Art Edge Models 'les Ministraux'
Mistral AI has announced two new edge models, 'Ministral 3B' and 'Ministral 8B', marking the first anniversary of Mistral 7B. These models set new benchmarks in knowledge, commonsense, reasoning, func
Mistral AI Launches Mathstral Model for Mathematical Reasoning
Mistral AI has announced Mathstral, a 7B model designed for mathematical reasoning and scientific discovery, in celebration of Archimedes' 2311th anniversary. The model features a 32K context window a
Mistral AI Launches New AI Endpoints in Early Access
Mistral AI is offering new AI endpoints in early access. There are three chat endpoints and one embedding endpoint, each with different performance and price trade-offs. Notably, Mistral-tiny is cost-
Introducing Customization Features from Mistral AI
Mistral AI has introduced new model customization features, allowing users to easily tailor Mistral's AI models to their specific needs, reducing costs and required expertise. With the new API and SDK
Mistral AI Unveils Its First Code Model, Codestral
Mistral AI has announced Codestral, its first AI model designed for code generation. This model supports over 80 programming languages and assists developers in writing code. Codestral saves time and
New Codestral 25.01 Released, Enhancing Coding Productivity
Mistral AI has announced the new Codestral 25.01 model. It generates and completes code about twice as fast as its predecessor. Supporting over 80 programming languages, it significantly boosts produc
Mistral introduces cost-effective batch API for AI developers
Mistral has announced a cost-effective batch API for AI developers, allowing for 50% lower costs compared to traditional synchronous API calls. It efficiently processes high-volume requests, enabling
Announcing Mistral 7B: The Most Powerful 7B Model Yet
Mistral AI has announced Mistral 7B, a language model with 7.3B parameters available under the Apache 2.0 license for unrestricted use. Mistral 7B outperforms the Llama 2 13B chat model and shows supe
Introduction of Remote Coding Agents Powered by Mistral Medium 3.5
Mistral Medium 3.5 has been announced, introducing remote coding agents in Vibe. This allows coding tasks to run independently in the cloud, notifying users upon completion. Additionally, a new Work m
Elevated Error Rates at Claude.ai Resolved
Claude.ai recently experienced elevated error rates, but the issue has been resolved following an investigation. The team is currently monitoring the recovery status. Such errors can affect user exper
DeepSeek API Introduces Disk-Based Context Caching
The DeepSeek API introduces disk-based context caching, allowing for the detection of duplicate user inputs to reduce service latency and cut costs by up to 90%. Users can expect quicker responses eve
DeepSeek-R1-Lite-Preview is now live: unleashing supercharged reasoning power!
DeepSeek has launched the new DeepSeek-R1-Lite-Preview. This model demonstrates high performance on AIME and MATH benchmarks, offering a transparent thought process in real-time. Open-source models an
Sparser, Faster, Lighter Transformer Language Models
The new transformer language models are characterized by their sparser, faster, and lighter design. This enhances computational efficiency, making them applicable to a wider range of applications, esp
GPT-5.5 Instant: Smarter, Clearer, and More Personalized Responses
GPT-5.5 Instant updates ChatGPT's default model to provide smarter and more accurate responses. It reduces hallucinations and improves personalization controls, allowing users to receive more tailored
Introducing MRC: A Networking Protocol for Large Scale AI Training
OpenAI has introduced MRC (Multipath Reliable Connection), a new networking protocol for supercomputers aimed at improving resilience and performance in large-scale AI training clusters. Released via