Elorian: Glimpse is a text model from Elorian, released on 2026-10-17. Tencent: Hy4 preview is a text model from Tencent, released on 2026-08-28. Both are served through api14 as chat models with text output, so the basic shape of a request is the same for either one.
The clearest split is in context and input modalities. Hy4 preview accepts text only, but its context window is 1048576 tokens, four times the 262144 tokens Glimpse allows. Glimpse accepts text, images and video as input, so it is the only one of the two that can look at a screenshot, a diagram or a clip directly. On parameters the two overlap heavily: both support tools, tool_choice, structured_outputs, response_format, reasoning, include_reasoning, logprobs, top_logprobs, temperature, top_p, top_k, min_p, frequency_penalty, presence_penalty, repetition_penalty, logit_bias and stop. Hy4 preview adds max_completion_tokens, parallel_tool_calls, reasoning_effort and seed. Glimpse adds max_tokens in the same role and nothing else exclusive. Neither has a capability the other completely lacks, so the decision comes down to context size, media input and price.
For most day to day work, pick based on the input you have. If your task is pure text and the prompt is very large, such as a long codebase, a stack of documents or a long conversation history, Hy4 preview is the natural choice because its context window is the larger of the two. If your task involves an image or a video, Glimpse is the only option, since Hy4 preview takes text only. For ordinary sized text prompts, both are viable, and the tie in tool and structured output support means you can swap between them without rewriting your tool schemas or your JSON response format. On price, Glimpse is the cheaper of the two for input, while Hy4 preview is the cheaper of the two for output, so the cheaper model overall depends on your input to output ratio. Check the live price table on this page for the current numbers.
Calling either model through api14 works the same way, because the API is OpenAI compatible. You point your client at the api14 base URL, send your key, and set the model field. To use Glimpse, set the model id to "elorian/glimpse". To use Hy4 preview, set it to "tencent/hy4-preview". Everything else in the request body stays as it is: messages, temperature, tools, response_format and the rest. If you already have code written against the OpenAI SDK, the change is a single string. This also makes it easy to route per request, sending media prompts to Glimpse and very long text prompts to Hy4 preview from the same application.
There are cases where the other model is still the better pick. If you are working with text only and your prompt fits comfortably inside 262144 tokens, Glimpse gives you the same tool and structured output support with cheaper input, which matters when you send a lot of context on every call. Conversely, if you need a reproducible run, Hy4 preview is the one that accepts a seed, and if you want to tune how much reasoning the model spends, it is also the one with reasoning_effort. Glimpse has no equivalent for either. So the choice is rarely about raw capability and more about which of these specific differences your workload actually touches.