Coding with AI / 4x RTX PRO 6000
4x RTX PRO 6000
Memory: 384 GB.
Memory bandwidth: 7168 GB/s.
Generation speed is bound by the chip's bandwidth; memory decides which models fit.
5 10 40 80 160 320 tok/s Opus 4.6 65.4% Sonnet 4.6 67% ↕ Comfort 20 ↕ Fast 60 GLM-4.7 41% · ~60 tok/s smart and fast GLM-4.7 41% on Terminal-Bench 2~60 tok/s · estimated How to run GLM-4.7 smart
Score: 41% (Terminal-Bench 2),
source
Speed: 60 tok/s (estimate: community-measured sibling GLM-4.6 355B-A32B AWQ ~60-65 tg single (vLLM TP4 на 4x RTX PRO 6000, huggingface.co/bullpoint/GLM-4.6-AWQ): та же архитектура/размер; квант приведён к декларированному Q4_K_M) Conditions: engine vLLM, quant Q4_K_M, context 128 Min memory: 238 GB Command: vllm serve glm-4.7
Every number above is either a measurement with a source or a labeled estimate with a method.
Unlabeled estimates do not exist on this site.
Ceiling timeline The strongest model fitting 4x RTX PRO 6000 (384 GB) at its release date, by SWE-bench Verified.
2025-07-22 Qwen3-Coder-480B 69.6% 2025-12-09 Devstral 2 72.2% 2025-12-22 GLM-4.7 73.8%