GLM-5.3 hits frontier range; Cerebras claims 30x GPU inference speed
创业邦 · wechat · 2026-08-20
Key items: Korean startup Motif trained Motif3 on $25M of government-backed Nvidia B200 GPUs, scoring 47 on the Artificial Analysis Intelligence Index—level with Alibaba's Qwen3.7 Max—while exploring a token-as-a-service model. Zhipu's GLM-5.3 API went live, scoring 60 on the AA index to enter the global frontier range, tied with Kimi K3 as top open model; pricing unchanged, weights open-sourcing next Friday. Cerebras unveiled CS-4 with three WSE-3 Turbo wafer-scale engines at 750 PFLOPS, claiming 4,400+ tokens/sec/user on GPT-OSS-120B—up to 30x GPU solutions—shipping Q3.
Also: OpenAI's Greg Brockman said the largest planned frontier RL training run is paused for two weeks to strengthen safety monitoring; multiple US states (PA, NY, TX) are tightening data center approvals; Google Cloud says it is automating part of its forward-deployed engineers' enterprise data governance work with AI.
More from Models
- Model at capacity for a straight week drives frustrated dev back to Opus — joonasvirtanen · 2026-10-16
- Decision models match LLMs on routing and safety — and are 7-8x faster — randal_olson · 2026-10-16
- Xiaomi MiMo V2.6 Pro takes #1 open-model spot on Agent Arena at $0.11/task — arena · 2026-10-16
- ChatGPT's 'best AI video tool' answer points to Runway integration, sparking ad doubts — tlakomy · 2026-10-16
- Microsoft's MAI finishes in-house 3T-parameter pretrained base models, post-training results coming — soumitrashukla9 · 2026-10-16
- Hands-on with Gemini 4 Argon: faster than Opus 5.5 but hallucinates more and costs more — prasenx · 2026-10-16