Baseten ships GLM-5.3 Fast: speed-optimized open-weight model for real-time workloads

baseten · x · 2026-09-03

Baseten released GLM-5.3 Fast, a speed-optimized version of Z AI's GLM-5.3 aimed at real-time workloads needing consistent performance at higher TPS.

• Model: 753B-A40B MoE, same 744B-A40B base as GLM-5.2; all gains come from scaled post-training on diverse, realistic task environments (full codebases, docs, testing tools, multi-step workflows)
• Capabilities: Z AI's strongest coding model yet — Terminal-Bench 3.0 jumps from 4.6% to 28.3% over GLM-5.2, plus notable gains in vulnerability discovery and security analysis
• Pricing: $2.10/M input tokens, $0.21/M cached input, $6.60/M output
• Access: OpenAI-compatible API on inference.baseten.co, with three thinking effort levels (max recommended for coding)

Original post →

More from Models

• User claims Gemini analyzed his tennis match video and improved his game — zacharynado · 2026-10-21
• Opus 5.5 tops OpenRouter spending in 20 of 29 task categories — mmurph · 2026-10-21
• Meta team builds open-source 30B LLM from scratch in months, hails 'second era' — TimDarcet · 2026-10-21
• DFlash drafter for Muse Glimmer-30B: 5 layers, blockwise 16-token spec decoding — TimDarcet · 2026-10-21
• EvoPress quantization shrinks 30B model from 56GB to 19GB at near-bf16 accuracy — TimDarcet · 2026-10-21
• Radar chart: Muse Glimmer-30B benchmarked against Gemma4-31B and Qwen3.6-27B — TimDarcet · 2026-10-21

Models channel →