Grok Code Fast 1 (xAI news) Aug 28, 2025
grok-code-fast-1
grok-code-fast-1 ranks first on Refusals + System Jailbreak (xAI).
Released August 28th, 2025 Compare with others
32
Benchmarks with a score
1
Ranked first, out of five or more models
3
In the top three
Unknown lab Aug 2025
grok-code-fast-1
No Epoch Capabilities Index score yet
- Terminal-Bench 2.0 25.8% #36 /43
- Toolathlon 18.5% #40 /50
- Terminal-Bench 25.8% #12 /18
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 70.8% | #72 of 155 | Lab model card | |
| IOI (Vals) | 4.33% | #43 of 55 | Official leaderboard | |
| DesignArena 3D Design | 1,048 | #109 of 136 | Official leaderboard | |
| WeirdML | 35.1% | #94 of 112 | Official leaderboard | |
| CodeClash | 1,006 | #7 of 8 | Official leaderboard | |
| DesignArena UI Component | 981 | #122 of 139 | Official leaderboard | |
| DesignArena Website | 989 | #134 of 151 | Official leaderboard | |
| DesignArena Game Dev | 978 | #128 of 143 | Official leaderboard | |
| DesignArena Code | 989 | #131 of 143 | Official leaderboard | |
| DesignArena Agentic Overall | 1,007 | #68 of 74 | Official leaderboard | |
| ALE-Bench | 853 | #118 of 126 | Official leaderboard | |
| DesignArena Full Stack | 988 | #46 of 49 | Official leaderboard | |
| Code Arena | 1,164 | #126 of 128 | Official leaderboard | |
| WebDev Arena | 1,165 | #104 of 105 | arena.ai, via Epoch AI | |
| DesignArena React Native | 1,028 | #56 of 56 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SnakeBench | 28 | #52 of 188 | Official leaderboard | |
| MCP-Universe | 26.4% | #11 of 30 | Official leaderboard | |
| MCPMark | 20.5% | #16 of 28 | Official leaderboard | |
| CompileBench | 73.3% | #12 of 20 | Official leaderboard | |
| Terminal-Bench | 25.8% | #12 of 18 | tbench.ai, via Epoch AI | |
| Pencil Puzzle Bench | 0% | #37 of 53 | Official leaderboard | |
| Toolathlon | 18.5% | #40 of 50 | Official leaderboard | |
| Terminal-Bench 2.0 | 25.8% | #36 of 43 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Refusals + System Jailbreak (xAI) | 0% | #1 of 5 | Lab model card | |
| General refusals (xAI) | 0% | #2 of 8 | Lab model card | |
| Refusals + User Jailbreak (xAI) | 0% | #2 of 5 | Lab model card | |
| AgentHarm | 17% | #4 of 6 | Lab model card | |
| AgentDojo | 26.9% | #6 of 8 | Lab model card | |
| BioLP-bench | 19.9% | #8 of 9 | Lab model card | |
| Virology Capabilities Test (VCT) | 28.7% | #30 of 31 | Lab model card | |
| MASK | 28.1% | #19 of 19 | Lab model card | |
| Cybench | 22.5% | #18 of 18 | Lab model card |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.
Its system card
Every result in the longer documents the lab published with it. 10 of them aren't in the launch post.
Grok Code Fast 1 Model Card Aug 26, 2025
Safety
Robustness
Alignment
- MASK 28.1
Bio and chem risk
- BioLP-bench 19.9
- Virology Capabilities Test (VCT) 28.7
- WMDP
Cyber risk
- Cybench 22.5