35 results for “inference”
Anthropic confirmed it is hiring a custom silicon team to co-design inference accelerators for Claude, joining Google, Amazon, Meta, Microsoft, and OpenAI in building proprietary AI chips. Samsung has been reported as a potential manufacturing partner.
Read more →
Developer slvDev runs a 28.9M-parameter LLM at 9.88 tokens/sec on a $10 ESP32-S3 microcontroller using 4-bit quantization and Per-Layer Embeddings from Google's Gemma 3n to fit the model in flash storage.
Read more →
Tsinghua-led researchers publish a cognitive-map AI framework in Nature Machine Intelligence that plans adaptively on neuromorphic chips, using 20-watt brain-like efficiency instead of megawatt GPU clusters.
Read more →
Nvidia open-sources cuFile and launches Storage-Next with 40+ vendors, introducing SCADA to let GPUs pull 512-byte AI inference data directly from SSDs — making flash a viable memory tier for KV caches and long contexts.
Read more →
Nvidia will invest up to $3 billion in Lancium, the power infrastructure developer behind the Stargate AI campus in Texas, taking a 20% stake and securing priority access to grid capacity for its AI customers.
Read more →
OpenAI has paused development on its upcoming Astra model after internal tests showed it reached a critical cybersecurity threshold — capable of independently finding and exploiting zero-day vulnerabilities in hardened systems. The model also solved 10 major open math problems.
Read more →
AMD, Supermicro and Spectro Cloud launch AMD Instinct Coder, a turnkey enterprise AI coding platform combining MI325X GPUs, EPYC CPUs, Pensando networking and Spectro Cloud's PaletteAI Inference Launchpad to route coding tasks locally via GLM-5.2 MoE model and selectively to frontier models, claiming up to 70% token cost reduction.
Read more →
OpenAI gives ChatGPT free users unlimited text chats and upgrades them to GPT-5.6 Luna, while Plus and Pro subscribers get a more reliable Sol model and a thinking slider.
Read more →
DeepSeek announced a significant API price increase on August 6, ending its ultra-low-cost strategy that disrupted the AI market. The hike follows surging demand, service instability, and a planned 1 GW data center build-out.
Read more →
Cloudflare open-sources Cloudflare OS, an AI agent workspace that gives every employee a secure, context-aware environment to build custom apps, automate workflows, and connect to internal systems — all governed by a capability-based security model called Gatekeepers.
Read more →
Liquid AI releases the LFM2.5 family — 1.2B and 2.6B open-weight models optimized for on-device agentic AI with hybrid architecture, multi-stage RL post-training, and day-one support across llama.cpp, MLX, vLLM, and NPU partners AMD and Nexa AI.
Read more →
ByteDance launches SeedRealtime, a native audio-visual full-duplex LLM that unifies audio, video, and text in a single architecture for real-time omni-modal interaction, now fully deployed at scale.
Read more →