Zhipu
19 articles tagged with this topic
Zhipu GLM-5.3 Scores Higher Without New Architecture — Chinese LLMs Turn Inward
Zhipu's GLM-5.3 hits stronger benchmarks with architecture identical to 5.2 — a training-only upgrade worth watching while peers chase new designs.
Zhipu GLM Runs Locally on Mac — And This Matters More Than It Looks
ds4 (co-maintained by Redis creator antirez) added Zhipu GLM Flash support this week, running on 128GB M4 Max. A concrete step for local Chinese LLMs.
TensorSharp Doubles Llama.cpp Decoding Speed, Could Halve LLM Deployment Costs
GLM-5.3-Flash hits 2x llama.cpp decoding on new TensorSharp framework; local deployment costs may drop sharply, but ecosystem maturity is unproven.
Zhipu Slashes AI Coding to 4 Cents—Price Isn't News, the Threshold Shift Is
GLM-5.3 Flash (Zhipu's Ox-Alpha) is open-source. 4 real coding tasks cost 4 cents RMB—~300x cheaper than Claude Opus 4.8. The threshold moved from 'us
Zhipu Releases GLM-5.3 Weights Tomorrow — Open Source Is China's AI Default
Zhipu releases GLM-5.3 weights tomorrow. China's top AI labs now treat open source as default — reshaping enterprise deployment costs and developer ec
Zhipu GLM Flash Beats Qwen — Size Doesn't Cut It
GLM Flash beat Qwen on three of four Reddit benchmarks. Qwen only led on graduate-level science by half a point — despite GLM being the larger model.
Zhipu flagship priced at 1/40 of Opus 4.8 — Chinese AI rewrites the default
GLM-5.3-Flash ties Opus 4.8 at 57 on Artificial Analysis, priced at 1/40th. First Chinese model combining frontier performance, low cost, and MIT lice
Zhipu Puts GLM-5.3-Flash on Hugging Face — China's Open-Weight Push Continues
Zhipu posts GLM-5.3-Flash to Hugging Face, betting on speed and low cost. As open-weight models multiply, can pay-per-call APIs survive?
Zhipu Open-Sources GLM-5.3-Flash: Nears Claude Opus at One-Tenth the Price
Zhipu open-sources GLM-5.3-Flash: 320B-param MoE with 18B active, claims near-Claude Opus 4.8 performance at one-tenth prior pricing under MIT license
Zhipu Open-Sources Ox Alpha vs DeepSeek — Open Source Becomes China's LLM Default
Z.AI confirms Ox Alpha is GLM's next gen, open-sourced tonight. After DeepSeek, a second Chinese AI firm bets open source for ecosystem.
GLM 5.3 Flash Stays a Rumor; Zhipu's Next Model Path Is Unclear
Zhipu's GLM 5.3 and a rumored "Flash" variant fuel speculation, but weights are unreleased. Rumors hint at open-source competition, not launch facts.
17K Attack Logs to a Chinese Model: Open-Source AI Defense Hits Reality
Hugging Face reportedly fed 17K+ attack logs to China's GLM 5.2. The platform now links papers, models, and executable infrastructure — boundaries are
Zhipu GLM's 'Hilarious' Thinking Goes Viral as Chinese Open LLMs Race on Inner Monologue
Zhipu GLM 5.3's 'hilarious' thinking hit r/LocalLLaMA. The meme masks Chinese open LLMs selling transparent reasoning as a post-R1 differentiator.
llama.cpp Fork Turns Retired AMD Server GPUs into AI Inference Rigs
Reddit user milpster and Zhipu AI's GLM team release a llama.cpp fork optimized for AMD GFX906, letting Mi50, Mi60, and Radeon VII run LLMs locally.
Open-Source Maintainer Uses AI to Unearth 281-Day Crash Bug
Maintainer built a single-file HTML triage tool with TRAE Work AI; a 281-day-old, zero-reply crash bug resurfaced across Gitee, GitHub, AtomGit.
Zhipu ships GLM-5.3 and ZCode same week — LLMs pivot from tokens to ecosystems
Zhipu shipped GLM-5.3 and ZCode the same week with a 2.39% pass-rate edge over Claude Code pairings. LLM firms now bundle models and tools into full e
GLM-5.3 Hits Artificial Analysis — China's First Open-Source Flagship Vetted
Zhipu's GLM-5.3 completes Artificial Analysis benchmarks — first Chinese open-source model to land in the global top tier with an independent score.
GLM-5.3 Matches GPT-5 at One-Third the Parameters — China Cracks Post-Training
Z.ai's GLM-5.3 hits GPT-5.6-Sol and Claude Fable 5 parity on Agent benchmarks with just 750B parameters. China's post-training bet is paying off.
Tim Dettmers Teases New Quantization Method: 7 tok/s on One Box — Industry Wary
Tim Dettmers claims GLM 5.3 hits 7 tok/s on a single DGX Spark. If true, enterprise LLM hardware costs halve — Reddit says: wait for benchmarks.