Top Stories
All Sections
Finance
About
14 results for "inference"
Google announced the TPU 8tx, its first chip with hardware-enforced confidential computing, and the TPU 8i Max, which fuses two chips for large-context inference — completing its eighth-generation lineup.
Thinking Machines Lab launched Inkling 2 Small, a 594-billion-parameter mixture-of-experts model optimized for coding, vision, and autonomous agentic tasks — entering the open-weight frontier AI race with an inference-efficient architecture.
Cloudflare patched a cross-request data exposure flaw in Workers AI that allowed one user's AI response to leak to another — arriving 24 hours after a separate AI inference security incident involving Alibaba's Qwen model.
Anthropic cut the price of its lightweight Claude Haiku model by 75%, matching OpenAI on cost and introducing a new Effort Setting that lets developers control compute trade-offs per request.
At Canada Fintech Forum, RBC and Cohere executives laid bare why AI that works in demos fails at mass customer scale: inference costs, regulatory review backlogs, and legacy systems not built for AI.
Google DeepMind's TurboQuant compresses AI inference memory 6x with zero accuracy loss and no retraining, delivering 8x faster throughput on H100 GPUs. It's already open source.
Fireworks AI, which helps enterprises deploy and run AI models at scale, is in active talks to raise a new funding round at a $15 billion valuation, with Index Ventures set to co-lead.
DeepSeek is permanently locking in a 75% discount on its V4-Pro API, setting new prices well below U.S. competitors and putting lasting pressure on the AI inference market.
Google has made Gemini 3.5 Flash generally available at $1.50 per million input tokens, offering frontier-level reasoning at four times the speed of comparable models — putting direct price pressure on OpenAI's fast inference lineup.
Factory AI inference speeds now allow vision-based defect detection to run inline on production stations, cutting floor space and cycle times while covering 100% of units produced.
Four Chinese AI labs — Z.ai, MiniMax, Moonshot, and DeepSeek — released open-weight coding models matching Western frontier performance at one-third the inference cost in May 2026.
Cloud provider Nebius Group is acquiring 20-person inference optimization startup Eigen AI for $643 million, signaling that making AI models run cheaper is now worth as much as raw compute capacity.
NVIDIA released Nemotron 3 Nano Omni, a 30B-parameter open model delivering 9x the throughput of comparable omni models — a direct play to own the agentic AI stack, not just the chips.
OpenAI is committing more than $20 billion to Cerebras over three years — doubling a prior arrangement — and taking a potential 10% equity stake in the chip startup as it builds an inference stack independent of Nvidia.