xAI
Grok API
xAI's Grok text models, both with a 500K-token context and image input.
2
моделей в линейке
The two Grok text models we serve both carry a 500K-token context and accept text and image input. 4.6 is the newer one, built for long-running multi-step agent tasks, and it is priced around seven times 4.5 on input — so unlike most families here, the newer model is the expensive one rather than the efficient one.
Коротко
- От
- 0.3 coins / 1 млн токенов
Модели этого семейства
| Модель | От | Режимы | Открыть |
|---|---|---|---|
| Grok 4.5xAI frontier model for coding, agentic tasks and knowledge work. 500K-token context, text and image input. | 0.3 coins/ 1 млн токенов | — | ОткрытьДокументация |
| Grok 4.6xAI upgrade over Grok 4.5, built for long-running multi-step agent tasks and interactive visual project creation. 500K-token context, text and image input. | 2.1 coins/ 1 млн токенов | — | ОткрытьДокументация |
Цены
| Модель | Вариант | Цена |
|---|---|---|
| Grok 4.5 | input | 0.3 coins/ 1 млн токенов |
| Grok 4.5 | output | 0.9 coins/ 1 млн токенов |
| Grok 4.5 | cacheRead | 0.075 coins/ 1 млн токенов |
| Grok 4.6 | input | 2.1 coins/ 1 млн токенов |
| Grok 4.6 | output | 6.3 coins/ 1 млн токенов |
| Grok 4.6 | cacheRead | 0.525 coins/ 1 млн токенов |
Что выбрать
Сильные стороны
- A 500K-token context on both tiers, including the cheap one.
- Image input on both tiers.
- Grok 4.5's output price is among the lowest for a model of its context length.
- Prompt caching at a quarter of fresh input on both.
Ограничения
- The newer 4.6 costs about seven times 4.5 on input, so upgrading is a real budget decision rather than a default.
- Only two tiers, with nothing between them.
- Cache reads are discounted to a quarter, not a tenth as in the GPT, Claude and Gemini lines.
Описание линейки от самого вендора: xAI — models