Meituan

Even Realities Raises $150M, Reaches $1B Valuation

Even Realities, a Shenzhen-based smart-glasses startup founded in 2023 by ex-Apple engineer Will Wang, raised $150 million in a pre-Series B round that values the company at $1 billion. Investors named in the round include Meituan and Tencent, alongside Chinese VC and PE backers such as CDH Investment, Monolith Management and CVC Capital; earlier rounds included Hong Kong’s Unicorn Capital Partners and Cyanhill Capital. The company sells display-first, camera-less AR glasses (Even G2) with a companion control ring (Even R1), emphasizes user privacy and has a majority of its users and roughly 80% of its developer community in the U.S. Industry data from IDC shows smart-glasses shipments surged and that Meta holds roughly 70% market share; IDC forecasts shipments could exceed 50 million by 2030. Earlier reporting in our database also notes Even Realities developed a proprietary optical system (Even HAO), has sold more than 10,000 pairs and expanded headcount substantially.

China demonstrates domestic AI training at scale

Multiple Chinese organisations in mid‑2026 reported large-scale model training runs using domestically produced processors, challenging the view that high-performance training still required Nvidia hardware. Meituan said it pre-trained LongCat‑2.0 (a trillion-parameter model) on a 50,000‑chip domestic cluster. A Huawei‑included research consortium reported full-parameter post‑training of DeepSeek‑V4‑Pro (1.6T MoE) on over 1,000 Ascend 910C chips, and Huawei’s Pangu Ultra runs and variants were also trained on large Ascend NPU clusters. Baidu reported training ERNIE 5.1 on a Kunlunxin-powered cluster, with industry reports linking that effort to Kunlun P800 chips and a claimed 97% effective training rate. Financial Times coverage (citing Morgan Stanley data) says some Chinese chips already outperform Nvidia’s H20 on several metrics. Together these datapoints indicate Chinese domestic chips have entered the AI training stack, though questions remain about training efficiency compared with best‑in‑class Nvidia solutions.

Anthropic launches Claude Sonnet 5; Fable/Mythos access restored

Anthropic released Claude Sonnet 5 as its new default mid-tier model with immediate rollout across Claude, Claude Code, API, Managed Agents and partner products. Sonnet 5 features a 1M-token context window, migration tooling, managed-agents platform updates, and list pricing of $3/M input and $15/M output with a temporary promotional rate of $2/M input and $10/M output. Third-party benchmarks show meaningful gains over Sonnet 4.6—especially on coding and agentic tasks—while some analyses reported higher token usage and task-level cost compared with Opus 4.8. Separately, Fable/Mythos 5 were approved for re-release after Anthropic's work with U.S. authorities. The release stimulated rapid ecosystem adoption but drew mixed reactions about naming, cost-efficiency, and absent frontier-model expectations.