#18
GLM 5.3 Flash
glm-5.3-flashHelpfulness
Instruction
Following
Comprehension
Empathy
Creative
Writing
Helpfulness
0.0
Instruction Following
0.0
Comprehension
0.0
Empathy
0.0
Creative Writing
0.0
Speed
Avg 90 tok/s
Overall Assistant Score
An average score combining the 5 main categories.
91.00 pts
Rank #18
83th Percentile
0
Novice
33
Capable
66
Proficient
100
Expert
GLM 5.3 Flash is Z.ai's first natively multimodal GLM-5 assistant: 320B-A18B MIT weights, 1M context, text/image/video in, at about one-tenth GLM 5.3's list price. Our tests put it just under 5.3 on helpfulness, instruction following, and comprehension for messy everyday work, while beating 5.2 Fast on those axes. Empathy stays cool and professional. Creative writing is still utilitarian GLM prose—clearer diffs than stories. Choose Flash when you want a fast, cheap, vision-capable helper; stay on 5.3 when the slower flagship should grind through the hardest multi-step finish.
Overall Score vs Price
Price axis is logarithmic (USD per 1M input tokens) · Higher score and lower price is better
Ringed point is this model
Anthropic
Baidu
Bytedance
DeepSeek
Google
Meta
Minimax
Mistral
Moonshot
OpenAI
Qwen
SpaceX AI
Tencent
Xiaomi
Zhipu AI
Compare this model
Intelligence
Overall Score · Higher is better
Speed
Output tokens per second · Higher is better
Price
USD per 1M input tokens · Lower is better
Zhipu AI Models
Overall Score · Same provider
GLM Family
Overall Score · Same model family
Closest Rivals
Overall Score · Nearest by overall rank