GLM-4.5 (Reasoning) vs Grok 4

Comparing 2 AI models · 6 benchmarks · Z AI, xAI

Most Affordable
Z AI logo
GLM-4.5 (Reasoning)
$0.57/1M
Highest Intelligence
xAI logo
Grok 4
87.7% GPQA
Best for Coding
xAI logo
Grok 4
55.1 Coding Index
Price Difference
5.2x
input cost range

Composite Indices

Intelligence, Coding, Math

Standard Benchmarks

Academic and industry benchmarks

Benchmark Winners

6 tests
Z AI logo

GLM-4.5 (Reasoning)

0

No clear wins

xAI logo

Grok 4

6
  • GPQA
  • MMLU Pro
  • HLE
  • LiveCodeBench
  • MATH 500
  • AIME 2025
Metric
Z AI logo GLM-4.5 (Reasoning)
Z AI
xAI logo Grok 4
xAI
Pricing
Per 1M tokens
Input Cost $0.57/1M $3.00/1M
Output Cost $2.19/1M $15.00/1M
Blended Cost 3:1 input/output ratio
$0.98/1M $6.00/1M
Specifications
Organization Model creator
Z AI xAI
Release Date Launch date
Jul 28, 2025 Jul 10, 2025
Performance & Speed
Throughput Output speed
50.8 tok/s 37.2 tok/s
Time to First Token (TTFT) Initial response delay
788ms 9172ms
Latency Time to first answer token
40177ms 9172ms
Composite Indices
Intelligence Index Overall reasoning capability
51.3 65.3
Coding Index Programming ability
43.3 55.1
Math Index Mathematical reasoning
73.7 92.7
Standard Benchmarks
GPQA Graduate-level reasoning
78.2% 87.7%
MMLU Pro Advanced knowledge
83.5% 86.6%
HLE Hard language evaluation
12.2% 23.9%
LiveCodeBench Real-world coding tasks
73.8% 81.9%
MATH 500 Mathematical problems
97.9% 99.0%
AIME 2025 Advanced math competition
73.7% 92.7%
AIME (Original) Math olympiad problems
87.3% 94.3%
SciCode Scientific code generation
34.8% 45.7%
LCR Code review capability
48.3% 68.0%
IFBench Instruction-following
44.1% 53.7%
TAU-bench v2 Tool use & agentic tasks
43.0% 74.9%
TerminalBench Hard CLI command generation
21.3% 37.6%

Key Takeaways

GLM-4.5 (Reasoning) offers the best value at $0.57/1M, making it ideal for high-volume applications and cost-conscious projects.

Grok 4 leads in reasoning capabilities with a 87.7% GPQA score, excelling at complex analytical tasks and problem-solving.

Grok 4 achieves a 55.1 coding index, making it the top choice for software development and code generation tasks.

All models support context windows of ∞