Hy4 Xploded vs GLM 5.3 Xploded
Tencent's Hy4 against Z.ai's GLM 5.3, both uncensored and both reading a million tokens with tool calling. Hy4 reads images and costs less for cached input; GLM 5.3 costs less for fresh input and output and writes longer replies.
| Hy4 Xploded | GLM 5.3 Xploded | |
|---|---|---|
| Model id | hy4-xploded | glm-5.3-xploded |
| Intelligence Index (base model) | - | 45 |
| Context | 1,048,576 tokens | 1,048,576 tokens |
| Longest reply | 65,536 tokens | 131,072 tokens |
| Tool calling | Yes | Yes |
| Images | Yes | No |
| Reasoning | On by default, adjustable | Always on |
| Input, per 1M tokens | $3.20 | $2.00 |
| Cached input, per 1M tokens | $0.60 | $1.00 |
| Output, per 1M tokens | $6.40 | $5.00 |
| A 25-turn agent session | $0.90 | $0.99 |
Intelligence Index is the Artificial Analysis Intelligence Index v4.3.2 for each base model; uncensored builds are not benchmarked separately. The agent session is about 750,000 input tokens re-read across 25 turns, 85% of them cache hits, and 25,000 tokens written. Prices are what MagmaRouter charges, checked 18 October 2026; the model pages always carry the current figures.
Use Hy4 Xploded when
- The job involves screenshots, diagrams or any other image. Hy4 Xploded reads them; GLM 5.3 Xploded is text only.
- The prompt is mostly re-read context, as an agent's is. Cached input is $0.60 per million tokens against $1.00.
- You want a second model family on the same task, to check one model's answer against another's.
Use GLM 5.3 Xploded when
- The reply is long. GLM 5.3 Xploded writes up to 131,072 tokens in one reply against 65,536.
- Fresh input and output dominate the bill: $2.00 in and $5.00 out per million tokens against $3.20 and $6.40.
Trying both
Both answer on https://api.magmarouter.com/v1 and the Anthropic-compatible endpoint with the same key, so the quickest comparison is your own prompt sent to each. The request is identical apart from the model id; the DeepSeek V4.1 Flash quickstart has it in curl, Python and JavaScript.
FAQ
What is Hy4 Xploded?
An uncensored build of Tencent's Hy4 flagship, model id hy4-xploded: 1,048,576 tokens of context, tool calling, reasoning and image input.
Does GLM 5.3 Xploded read images?
No. GLM 5.3 Xploded is text only; Hy4 Xploded reads images.
Can I switch between them?
Yes. Same endpoint, same key: change the model id from hy4-xploded to glm-5.3-xploded or back.