GLM-5.3 lands on Baseten: Terminal-Bench 3.0 jumps from 4.6% to 28.3%
baseten · x · 2026-09-05
Zhipu's GLM-5.3 is live exclusively on Baseten Model APIs with vision support, converting images to code. It runs on the same 744B-A40B MoE base as GLM-5.2 (753B-A40B total), priced at $1.40 input / $0.14 cache / $4.40 output per 1M tokens.
All gains come from scaled post-training on realistic environments—full codebases, documentation, testing tools, and multi-step workflows. Terminal-Bench 3.0 jumps from 4.6% to 28.3% over GLM-5.2, with notable gains in vulnerability discovery. Three thinking effort levels are offered; max is recommended for coding.
Related event: Zhipu's GLM-5.3 Launches Exclusively on Baseten with Big Coding Gains(2 posts)→
More from coding & agent
- The Inverted Prompt: A Satirical Guide to Resume-Driven Over-Engineering — adrianscottcom · 2026-09-05
- Together AI Ships Guide to Deploy a Chat API on Render Without Kubernetes — togethercompute · 2026-09-05
- GPT-6 Astra lands in Netlify AI Gateway and Agent Runners with zero config — thisiskp_ · 2026-09-05
- Dev tests fal.ai generative platform: streaming blocked by RTC issues, but Reactor is fun — flngr · 2026-09-05
- GPT-6 Astra base instructions run 16% longer than GPT-5.6 Sol's, cutting permission loops — WolframRvnwlf · 2026-09-05
- Entire proposes repo mirroring to stop agents re-cloning and hide team context — craigsdennis · 2026-09-05