RobinsonLabshomelab · ai · ops
CLUSTER · 16 GPU NODES · 40G RDMA
The archive
News · edition

Edition, Mon, 03 Aug 2026.

3 stories on the wire this day. Open any story for its sources. Summaries are written by software from the linked reporting and can be wrong. Spot an error? Email [email protected] and we will correct or remove it.

AI & Technology

London-based OLIX raised $312 million at a $3.3 billion valuation, more than tripling its February valuation, as it develops optical computing chips designed to make AI inference more efficient and affordable.

2 sources
Lead
this edition
Still developing
Anthropic debuts Mythos-class AI with record coding and science leapsClaude Fable 5 and restricted Mythos 5 set new benchmarks, solving a 50-million-line code migration in a day and generating novel drug candidates autonomously.The Decoder
OpenAI unveils Jalapeño, its first custom inference chip designed with BroadcomThe new processor aims to optimize performance-per-watt for large language model inference, reducing reliance on Nvidia GPUs.TechCrunch
Moonshot AI's Kimi K3 becomes world's largest open-weight model, tops code benchmarksThe 2.8-trillion-parameter model released July 16 ranked first in the Frontend Code Arena with 1,679 points, beating Claude Fable 5 in blind developer testing.Tom's HardwareArena.ai
ServiceNow hits $1B AI ACV and raises full-year outlook in strong Q2 2026 earningsServiceNow reported Q2 2026 revenue of $3.987B (up 24% YoY) and announced its AI platform crossed $1B in annual contract value, prompting a raised full-year subscription revenue forecast.Outlet
Meta raises 2026 AI infrastructure spending floor to $130 billion as stock dips on ROI questionsCEO Mark Zuckerberg cited confidence in the investment despite market pushback on the massive capital expenditure increase.Fortune
SpaceXAI launches Grok 4.5 as an Opus-class model at lower costThe new model excels at coding, agentic tasks, and knowledge work with twice the token efficiency of competitors.SpaceXAITechCrunch
xAI releases Grok 4.5 with 1.5 trillion parameter MoE architecture and competitive pricingThe new model launched July 9 alongside GPT-5.6, featuring training on Cursor data and achieving 83.3 percent on Terminal-Bench 2.1 while using approximately 25 percent fewer output tokens than Opus 4.8.AIapps
Research firm Artificial Analysis found DeepSeek's V4-Flash costs roughly 3 cents to run through a full benchmark suite, making it by far the least expensive well-known model compared to 86 cents for Kimi K3, $1.86 for GPT-5.6 Sol, and $3.15 for Claude Fable 5.

AI & Technology

AI & Technology
DeepX, a South Korean edge-AI chip designer, hit a valuation of $2.2 billion (approximately 3.14 trillion Korean won) following its Series D round, representing roughly a fourfold increase from its previous valuation.
1 source
AI & Technology

UK photonic-chip startup OLIX triples valuation with $312M Series B

London-based OLIX raised $312 million at a $3.3 billion valuation, more than tripling its February valuation, as it develops optical computing chips designed to make AI inference more efficient and affordable.

London-based OLIX raised $312 million at a $3.3 billion valuation, more than tripling its February valuation, as it develops optical computing chips designed to make AI inference more efficient and affordable.

The round was led by Fundomo with participation from Arm, Hudson River Trading, and Netflix co-founder Reed Hastings, while existing investors including the UK government's Sovereign AI fund also increased their commitments. OLIX appointed Professor Nick McKeown to its board and Matt Briers as CFO, and expects to deliver its first chip, the DX-1 decode accelerator, to customers in the second half of 2027. The company's architecture uses specialized chips for different stages of token production connected by an optical interconnect, aiming to break the efficiency limits of general-purpose GPU-based inference.

2 sources
OLIX
Tech.eu
AI & Technology

DeepSeek V4-Flash emerges as cheapest major AI model at 3 cents per test

Research firm Artificial Analysis found DeepSeek's V4-Flash costs roughly 3 cents to run through a full benchmark suite, making it by far the least expensive well-known model compared to 86 cents for Kimi K3, $1.86 for GPT-5.6 Sol, and $3.15 for Claude Fable 5.

Research firm Artificial Analysis found DeepSeek's V4-Flash costs roughly 3 cents to run through a full benchmark suite, making it by far the least expensive well-known model compared to 86 cents for Kimi K3, $1.86 for GPT-5.6 Sol, and $3.15 for Claude Fable 5.

The model, launched alongside the heavier V4-Pro, is priced at $0.14 per million input tokens and $0.28 per million output tokens, designed for high-volume applications like chatbots and coding assistants. While V4-Flash scored 50 on the Intelligence Index—below the frontier models that scored around 59—the release continues a price war trend where Chinese labs are driving costs down significantly, challenging the economics of premium-priced Western models.

2 sources
The Next Web
Yahoo Finance
AI & Technology

Korean edge-AI chip designer DeepX reaches $2.2B valuation in Series D

DeepX, a South Korean edge-AI chip designer, hit a valuation of $2.2 billion (approximately 3.14 trillion Korean won) following its Series D round, representing roughly a fourfold increase from its previous valuation.

DeepX, a South Korean edge-AI chip designer, hit a valuation of $2.2 billion (approximately 3.14 trillion Korean won) following its Series D round, representing roughly a fourfold increase from its previous valuation.

The company raised 42 billion won from existing backers BNW Investment and DS Asset Management, and is reportedly in talks to raise up to 300 billion more by September. DeepX specializes in edge-AI chip design for efficient inference on devices, and this funding round reflects continued investor interest in specialized AI hardware beyond the data center GPU market dominated by NVIDIA.

1 source
AI Weekly
Every edition

The wire publishes daily and keeps an archive of every edition, listed by date, newest first. If a summary is wrong, we correct or remove it.

Back to the archive Latest edition