Head to head
Grok 4.7 vs GPT-6.1 Sol
Which is better, Grok 4.7 or GPT-6.1 Sol?
GPT-6.1 Sol finishes ahead of Grok 4.7 on the index, 90.3 to 84.5, and leads on all four scores: craft, speed, control and value. Both are scored on the same rubric, in the same week, and neither placing is sponsored.
Updated
Grok 4.7
23 of 73
84.5
xAI's new flagship: strong on benchmarks, cheap on output, and less loved by human voters.
GPT-6.1 Sol
04 of 73
90.3
Near-Astra answers at a fifth of the price. The OpenAI model most teams should deploy now.
The four scores, side by side
| Score | Grok 4.7 | GPT-6.1 Sol | Difference |
|---|---|---|---|
| Craft | 84 | 92 | +8 |
| Speed | 82 | 82 | — |
| Control | 86 | 91 | +5 |
| Value | 86 | 89 | +3 |
| Index score | 84.5 | 90.3 | +5.8 |
Which to pick
GPT-6.1 Sol is the safer pick on every axis we score. Grok 4.7 earns its place in the index, but nothing in these four numbers argues for it over GPT-6.1 Sol.
Grok 4.7
Grok 4.7 arrived on 21 September and scores well on the Artificial Analysis index, level with Xiaomi's MiMo V2.6 Pro and ahead of GPT-5.6 Sol. Output is unusually cheap at six dollars per million tokens, and it keeps the family's habit of answering the question directly. Human voters on the arena rank it far lower than its benchmarks, below its own predecessor, so test it on your own prompts before you commit.
Where it shines
- Six dollars per million output tokens, low for this tier
- Half a million tokens of context
- Answers directly rather than hedging
GPT-6.1 Sol
GPT-6.1 Sol launched at DevDay a week after GPT-6 Sol and replaces it at the same price. OpenAI describes it as near-Astra performance at a lower cost, and the leaderboards bear that out: it scores just below Astra on the Artificial Analysis index and well above it with human voters. At two dollars in and ten out, it is priced head to head with Claude Sonnet 5.5, and it is the default choice inside the OpenAI ecosystem.
Where it shines
- Close to Astra on benchmarks for a fifth of the price
- Preferred over Astra by human voters
- Every framework and agent toolkit supports it from day one