Grok 4.5 vs Hy3 preview
Price, context and benchmark scores side by side — every number from the live model catalogue.
| Grok 4.5x-ai | Hy3 previewtencent | |
|---|---|---|
| Input / 1M tokens | $2.00 | $0.063 |
| Output / 1M tokens | $6.00 | $0.210 |
| 1M in + 300K out | $3.80 | $0.13 |
| Context window | 500K | 262K |
| Intelligence index | 53.8 | 41.2 |
| Coding index | 72.4 | 58.8 |
| Agentic index | 45.7 | 30.7 |
| Cheapest provider | xAI | GMICloud |
| Providers serving it | 1 | 1 |
| Weights | Closed | Open |
| Image input | Yes | No |
| Tool calling | Yes | Yes |
| Reasoning mode | Yes | Yes |
| Released | Jul 8, 2026 | Apr 22, 2026 |
Which one to pick
Pick Grok 4.5 if…
- You want the higher general-intelligence score (53.8).
- Code quality is the priority — it leads on the coding index (72.4).
- You are running agent loops — higher agentic index (45.7).
- You need the bigger context window (500K).
Pick Hy3 preview if…
- Token cost matters: $0.210 per 1M output vs $6.00.
- You want the option to self-host — weights are public.
FAQ
Is Grok 4.5 better than Hy3 preview?
Grok 4.5 scores higher on general intelligence (53.8 vs 41.2), while Hy3 preview is 29× cheaper per output token ($0.210 vs $6.00). For code, Grok 4.5 leads (72.4 coding index). Grok 4.5 takes the larger context at 500K.
Which is cheaper, Grok 4.5 or Hy3 preview?
Hy3 preview, at $0.210 per 1M output tokens versus $6.00. On a 1M input + 300K output workload that is $0.13 against $3.80.
Can I self-host Grok 4.5 or Hy3 preview?
Grok 4.5: closed weights, API only. Hy3 preview: open weights (tencent/Hy3-preview).