DataLearner logo

Gemma 4 31BvsGLM-5

Across 10 shared benchmarks, GLM-5 leads overall: Gemma 4 31B wins 3, GLM-5 wins 7, with 0 ties and an average score difference of -4.99.

DeepMind
Gemma 4 31B

DeepMind · 2026-04-02 · Chat model

智谱AI
GLM-5

智谱AI · 2026-02-11 · Chat model

Gemma 4 31B3 wins(30%)(70%)7 winsGLM-5

Benchmark scores

Grouped by capability, sorted by largest gap within each. 10 shared benchmarks.

Agent Level Benchmark

GLM-5 3/3
BenchmarkGemma 4 31BGLM-5Diff
τ²-Bench - Telecom65.50149 / 264Normal (With Tools)97.4016 / 264Normal (With Tools)-31.90
τ²-Bench76.9020 / 44Thinking (With Tools)89.704 / 44Thinking (With Tools)-12.80
Terminal Bench Hard30.30110 / 244Normal (With Tools)39.4053 / 244Normal (With Tools)-9.10

General Knowledge

GLM-5 2/3
BenchmarkGemma 4 31BGLM-5Diff
LiveBench61.6264 / 117Normal (No Tools)68.8543 / 117Normal (No Tools)-7.23
HLE11.80369 / 563Normal (No Tools) · Text only7.60427 / 563Normal (No Tools) · Text only+4.20
CritPt1.40135 / 200Thinking (No Tools)2127 / 200Thinking (No Tools)-0.60

General Evaluation

Gemma 4 31B 1/1
BenchmarkGemma 4 31BGLM-5Diff
GPQA Diamond76.30270 / 462Normal (No Tools)66.60349 / 462Normal (No Tools)+9.70

Instruction Following

GLM-5 1/1
BenchmarkGemma 4 31BGLM-5Diff
IF Bench53.50144 / 282Normal (No Tools)55.20132 / 282Normal (No Tools)-1.70

Long Context

Gemma 4 31B 1/1
BenchmarkGemma 4 31BGLM-5Diff
AA-LCR46.70141 / 170Normal (No Tools)43.70151 / 170Normal (No Tools)+3

Math and Reasoning

GLM-5 1/1
BenchmarkGemma 4 31BGLM-5Diff
AIME 202689.2025 / 29Thinking (No Tools)92.7018 / 29Thinking (No Tools)-3.50

Specs

FieldGemma 4 31BGLM-5
PublisherDeepMind智谱AI
Release date2026-04-022026-02-11
Model typeChat modelChat model
ArchitectureDenseMoE
Parameters30.7B744B
Context length256K200K
Max output32K128K

API pricing

Prices use DataLearner records when available; missing fields are not inferred.

ItemGemma 4 31BGLM-5
Text inputNot public$1 / 1M tokens
Text outputNot public$3.2 / 1M tokens
Cache writeNot public$0.2 / 1M tokens

One or both models have incomplete public pricing.

Summary

  • Gemma 4 31Bleads in:General Evaluation (1/1), Long Context (1/1)
  • GLM-5leads in:Agent Level Benchmark (3/3), General Knowledge (2/3), Instruction Following (1/1), Math and Reasoning (1/1)

On average across the 10 shared benchmarks, GLM-5 scores 4.99 higher.

Largest single-benchmark gap: τ²-Bench - Telecom — Gemma 4 31B 65.50 vs GLM-5 97.40 (-31.90).

Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.