xAI's Grok 4.3 and Moonshot AI's Kimi K2.6 compared on specs, pricing, and capabilities — as configured in AI Crucible ensembles.
| Specification | Grok 4.3 | Kimi K2.6 |
|---|---|---|
| Provider | xAI | Moonshot AI |
| API model ID | grok-4.3 | kimi-k2.6 |
| Description | xAI's flagship Grok model with 1M context, supporting both reasoning and fast non-reasoning modes. | Moonshot AI's latest flagship native multimodal agentic model with deep reasoning and 256K context. |
| Context window | 1M tokens | 262K tokens |
| Max output tokens | 32,768 | 8,192 |
| Input cost ($/1M tokens) | $1.50 | $1.14 |
| Output cost ($/1M tokens) | $3.00 | $4.80 |
| Cache read cost ($/1M tokens) | $0.24 | $0.19 |
| Latency | medium | medium |
| Reasoning model | Yes | Yes |
| Vision (image input) | Yes | Yes |
| Tool use | Yes | Yes |
See all model comparisons, full model specifications, benchmark results, or model comparison articles.