Grok 4.5
GrokBalanced
ThinkingTool UseVisionStructured Output
About this model
Grok 4.5 is a model from SpaceXAI with frontier performance on coding, knowledge work, and STEM.
Performance Tier
Balanced
Grok 4.5 is a balanced model from Grok : strong performance at a reasonable price.
Strong cost-performance ratio. Reliable for most professional use cases without premium pricing.
Pricing
This model is included in Elosia plansModerate
Moderate cost. A balanced choice for regular use.
| Type | per 1M tokens |
|---|---|
| Input (prompt) | $2.00 |
| Output (completion) | $6.00 |
| Cache read | $0.300 |
Capabilities
Context Length500K
Max Output Tokens450K
TokenizerGrok
Inputtext, image, file
Outputtext
Release DateJuly 8, 2026
Benchmarks
General Intelligence
MMLU
Not reported
Mathematics
MATH-500
Not reported
Programming
HumanEval
Not reported
Agentic
SWE-bench Pro
64.7%
Terminal-Bench 2.1
83.3%
Where does Grok 4.5 stand?
Compare its performance index against every other model.
Recommended Use Cases
CodingAnalysisResearchGeneral ChatMathematics
Strengths
- Strong on coding, agentic tasks and STEM knowledge work
- Competitive agentic coding: 83.3 on Terminal-Bench 2.1 and 64.7 on SWE-bench Pro (single-model)
- Notably token-efficient on agentic coding versus comparable flagships, at about 80 tokens/second (per xAI)
- 500K-token context window with multimodal (text, image) input, function calling, web and X search, and code execution
Limitations
- xAI published only coding and agentic benchmarks; no MMLU, GPQA, math or reasoning scores to gauge broader capability
- Trails the current SWE-bench Pro coding leaders (64.7 vs about 80)
- Smaller third-party tooling ecosystem than OpenAI or Anthropic
Frequently asked questions
Resources
This model may use your data for training
