9 results for “LLM”
DeepSeek V4 appears headed for an August 3 launch as SiliconFlow raises API prices tenfold and developers report new capabilities in phased API testing, while OpenAI cuts GPT-5.6 prices 80%.
Read more →
Tencent open-sources AngelSpec, a unified training framework that uses workload-aware strategies — multi-token prediction for dialogue and block diffusion for code and math — to boost LLM inference throughput across concurrency levels 4-64.
Read more →
Google's updated Gemini macOS app adds Fn-key voice dictation and screen-aware reasoning that can read multiple local files and insert AI output directly into any document — challenging Apple Intelligence and ChatGPT on the Mac desktop.
Read more →
Anthropic released Claude Opus 5, delivering near-Fable 5 performance at half the price ($5/$25 per million tokens) with new effort dial and mid-task switching controls.
Read more →
AMD launches Helios AI rack-scale system with 72 MI455X GPUs and 18 EPYC Venice CPUs, claiming 30% better tokens-per-dollar than Nvidia. OpenAI, Anthropic, Meta, and Microsoft commit to gigawatt-scale deployments starting Q4 2026.
Read more →
Hugging Face disclosed a July 2026 breach that hit internal datasets and service credentials. The company used GLM 5.2 on-prem after commercial AI guardrails blocked log analysis.
Read more →
Moonshot AI launched Kimi K3, a 2.8-trillion-parameter open 3T-class model with a 1-million-token context window and native multimodal training. The model is live on kimi.com and via API; full weights are scheduled for July 27, 2026.
Read more →
Meta is reportedly in talks to lease AI computing capacity to Anthropic in a deal worth up to $10 billion over two years, according to The New York Times, marking the latest cross-competitor compute arrangement driven by the AI hardware shortage.
Read more →
DeepSeek V4 Flash is an open-weight AI model focused on long context, low-cost inference, and practical coding and agent workflows.
Read more →