GPT-5.6: Frontier intelligence that scales with your ambition
OpenAI News / Jul 9, 2026
- トークン効率向上
- コスト当たり性能改善
- オンデマンド能力強化
Matched posts: 13
OpenAI and Broadcom unveil LLM-optimized inference chip
OpenAI News / Jun 24, 2026
- LLM専用アクセラレータ発表
- 設計から9か月でテープアウト
- 初期で高いperf/watt示唆
Growing the Cloudflare AI team with talent from Ensemble AI
Cloudflare / Jun 15, 2026
- NdLinearで構造を保った圧縮
- Workers AIでサーバレスGPU推論
- 推論コストとGPU利用率の改善
I/O 2026: Welcome to the agentic Gemini era
Gemini / May 19, 2026
- gemini 3.5 flash 高速かつ低コスト
- gemini omni flash 公開とAPI展開
- synthid と Content Credentials 拡張
Speeding up agentic workflows with WebSockets in the Responses API
OpenAI News / Apr 22, 2026
- Up to 40% faster agentic workflows
- WebSocket mode reuses in-memory response state
- Achieves 1,000+ tokens/sec for fast models
Building the agentic cloud: everything we launched during Agents Week 2026
Cloudflare / Apr 20, 2026
- Sandboxes GA for persistent agent runtimes
- Artifacts: Git‑compatible versioned storage
- Cloudflare Mesh + Managed OAuth for secure agent access
Unweight: how we compressed an LLM 22% without sacrificing quality
Cloudflare / Apr 17, 2026
- Lossless BF16 exponent Huffman coding
- Reconstructive matmul decompresses in shared memory
- 15–22% model size reduction (~3GB VRAM saved)
Cloudflare’s AI Platform: an inference layer designed for agents
Cloudflare / Apr 16, 2026
- 単一APIで複数プロバイダ
- 自前モデルをWorkers AIへデプロイ
- 自動フェイルオーバーと再接続
Building the foundation for running extra-large language models
Cloudflare / Apr 16, 2026
- Prefill–decode disaggregation cut tail latencies and improved intertoken throughput
- x-session-affinity prompt caching raised cache hit ratio from ~60% to ~80%
- Infire: multi-GPU support, lower memory overhead, and <20s cold starts
AI fundamentals
OpenAI News / Apr 10, 2026
- パターン認識ソフト
- LLMは次語予測
- 用途でモデル選択
Powering the agents: Workers AI now runs large models, starting with Kimi K2.5
Cloudflare / Mar 19, 2026
- Kimi K2.5: 256k context & multi-turn tool calling
- Prefix caching + x-session-affinity cuts prefill cost and improves TTFT
- Pull-based async API for durable, high-volume inference
Gemini 3.1 Flash-Lite: Built for intelligence at scale
Gemini / Mar 3, 2026
- Preview via Gemini API and Vertex AI
- Low cost: $0.25/1M input, $1.50/1M output
- 2.5× faster TTFAT and 45% faster output
- Run Whisper on-device with react-native-executorch
- Privacy and offline-first, sub-second latency
- Plan for model size, memory and battery trade-offs
Previous1 / 1Next