A September 9, 2026 analysis of twelve AI accelerators from NVIDIA, AMD, Google, AWS, Cerebras, and Tenstorrent finds that memory per dollar, not FLOPS, is the primary purchasing criterion for AI comp…
A developer published an AI agent evaluation playbook, a repeatable test battery for vetting whether models running inside agent frameworks can be trusted with semi-sensitive content and real write ac…
Cerebras added Alibaba's Qwen 3.8 27B to its public API, delivering 1,500 tokens per second, enabling a 300-word response in under half a second. The model scores 61.7% on SWE-bench Pro, ahead of Clau…
OpenAI's Astra launch has locked out paying Plus and Pro subscribers while enterprise customers received first access, prompting CEO Sam Altman to apologize on X for the 'messy' rollout. Separately, r…
Mistral announced Agentic Search, a feature enabling autonomous search workflows with a data opt-out guarantee for training. The company also highlighted inference optimization for agentic use cases, …
Andreessen Horowitz (a16z) announced an investment in Gimlet Labs, which it calls the first multi-silicon inference cloud designed to address the global shortage of AI compute power. The firm notes th…
OpenAI shipped GPT-6 Astra on Thursday at $10 per million input tokens and $50 per million output, 2.5 times the price of GPT-5.6 Sol, and rated it 'Critical' for cyber capability under its Preparedne…
Developer Tashfeen Ahmed released FreeLLMAPI, an open-source router that unifies 34 free LLM providers behind a single OpenAI-compatible endpoint. The tool encrypts provider keys locally, tracks rate …
The Institute of Foundation Models at Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) released K2 Horizon, a family of six open-source AI models ranging from 0.9 billion to 375 billio…
Meryem Arik, co-founder of inference company Doubleword, said most companies overpay for AI inference by 2x to 5x, sometimes by an order of magnitude, because their inference stacks do not match their…
OpenAI's August 13 limited API preview of Ultrafast for GPT-5.6 Sol reports up to 750 output tokens per second, as much as 14 times Standard speed, with a Cerebras benchmark showing a 5.59-times end-t…
AMD's general manager of its data center GPU business, Andrew Dieckmann, said the company's first rack-scale AI system, Helios, is designed to challenge Nvidia's dominance by offering openness and cus…
The Register reports that 100+ tech giants have warned that AI attacks are coming, but have skipped the part where they pay for defenses, highlighting the industry's role in creating the problem and s…
Nvidia and Cerebras are touting benchmark results at Hot Chips that show inference speeds of 3,400 tokens per second on Gemma 4 31B, but The Register argues these figures are unattainable in real prod…
Anthropic has filed confidentially for a potential IPO as soon as October and may pitch investors on revenue opportunities above $30 trillion, according to a Wall Street Journal report. Connor Group m…
Nvidia's Groq 3 LPX claims a throughput of 3,400 tokens per second, but achieving that requires a cluster of at least 64 accelerators, whereas Cerebras can reportedly achieve similar performance with …
Nvidia has moved its Groq 3 LPX inference chip into full production, claiming 3,400 tokens per second on Gemma 4 31B, which it says is four times faster than Cerebras. However, Nvidia requires at leas…
Nvidia's first benchmarks for its Groq 3-based LPX racks show a 4x speedup over the nearest alternative, delivering 3,400 tokens per second on Google's Gemma 4 31B model, according to independent test…
Nvidia faces rising competition in the AI data center processor market as AMD, Google, Cerebras, Amazon, Meta, and Microsoft invest in alternatives, with AMD securing 6 GW deployment commitments each …
AI vendors are increasingly turning to custom silicon from Cerebras and Marvell for inference workloads, and Waymo is designing its own chips for robotaxis, signaling a shift away from reliance on tra…