Grok 4.6 vs Step 5 Preview
Grok 4.6 
Step 5 Preview Verdict
- Grok 4.6 (medium) leads on quality: 66.2 vs 65.0.
- Grok 4.6 (medium) is stronger in agents & tools, coding, knowledge.
- Step 5 Preview is stronger in composite, long context, reasoning, multimodal.
- Step 5 Preview is 2.1× cheaper ($1.43 vs $3.00 per 1M blended).
| Metric | Grok 4.6 (medium) | Step 5 Preview |
|---|---|---|
| BenchLeader Index | 66.2 | 65.0 |
| Agents & tools score | 73.3 | – |
| Coding score | 64.1 | – |
| Composite score | 83.6 | 84.4 |
| Knowledge score | 77.6 | 72.0 |
| Long context score | 67.0 | 70.8 |
| Reasoning score | 67.7 | 82.1 |
| Multimodal score | – | 60.8 |
| Blended price $/M | $3.00 | $1.43 |
| Output speed | 57 tok/s | – |
| Time to first answer | 34.1 s | – |
| Context window | 500k | 1M |
| SciCode | 54.6% | – |
| AA Intelligence Index | 43.0 | 43.6 |
| AA-LCR | 81.0% | 88.3% |
| MMMU-Pro | – | 76.4% |
| AA-Omniscience | 28 | 16.4 |
| GPQA Diamond (AA) | 93.5% | – |
| Humanity's Last Exam (AA) | 42.1% | 46.5% |
| SciCode (AA) | 55.9% | 58.9% |
| ARC-AGI-1 | 87.5% | – |
| ARC-AGI-2 | 61.3% | – |
| CritPt | 17.7% | 20.9% |
| GDPval (AA) | 57.1% | 53.6% |
| τ²-Bench Banking (AA) | 44.3% | – |
| DeepSWE | 67.5% | – |
| CursorBench | 67.1% | – |
Data as of 2026-09-19. Best configuration of each model; every score links to its source on the model pages.
Grok 4.6 vs Step 5 Preview: questions
- Is Grok 4.6 better than Step 5 Preview?
- Grok 4.6 (medium) leads on quality: 66.2 vs 65.0. The BenchLeader Index combines every independent quality benchmark; Grok 4.6 (medium) is ahead overall as of 2026-09-19, but check the category scores for your use.
- Which is cheaper, Grok 4.6 or Step 5 Preview?
- Step 5 Preview is cheaper: $1.43 against $3.00 per million tokens, blended at three input tokens per output token.
- Which has the larger context window?
- Step 5 Preview accepts more context: 1M against 500k tokens.