Interactive Benchmark Matrix

Model Scatter Plot

Pick any two metrics for the X and Y axes. Every model is a floating point on the chart, so you can visually compare trade-offs between accuracy, speed, and token usage.

Throughput (tps)Overall Score (%)Gemma 4 31BGLM 4.7GLM 5.2 High (self)
Gemma 4 31B

Overall: 49.4%

Coding: 45.5%

Reasoning: 53.5%

Tokens / Task: 279.3

Provider: Cerebras

GLM 4.7

Overall: 41%

Coding: 32.3%

Reasoning: 50.3%

Tokens / Task: 284.7

Provider: Cerebras

GLM 5.2 High (self)

Overall: 64.9%

Coding: 57.5%

Reasoning: 72.9%

Tokens / Task: 0

Provider: Devin (self-benchmark)