Cohere: Command R (08-2024) vs Tencent: Hy3 preview
A side-by-side comparison of Cohere: Command R (08-2024) ($0.52 / 1M tokens) and Tencent: Hy3 preview ($0.63 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.
Specifications
| Cohere: Command R (08-2024) | Tencent: Hy3 preview | |
|---|---|---|
| Provider | Cohere | Tencent |
| Input | $0.52 / 1M tokens | $0.63 / 1M tokens |
| Output | $2.10 / 1M tokens | $2.10 / 1M tokens |
| Cache read | - | $0.21 / 1M tokens |
| Context window | 128,000 | 262,144 |
| Max output | 4,000 | 235,929 |
| Input | text | text |
| Output | text | text |
| Tool calling | ||
| Structured output | ||
| Reasoning control | ||
| Added | Aug 30, 2024 | Apr 22, 2026 |
Key differences
- Cohere: Command R (08-2024) is 17% cheaper on input tokens.
- Tencent: Hy3 preview reads 2.0× more context (262,144 tokens).
- Only Cohere: Command R (08-2024) supports structured JSON output.
- Only Tencent: Hy3 preview supports reasoning effort control.
- A request with 1,000 input and 500 output tokens costs $0.00158 with Cohere: Command R (08-2024) and $0.00168 with Tencent: Hy3 preview.
In depth
Cohere: Command R (08-2024) is a text model from Cohere, released on 30 August 2024 under the model id cohere/command-r-08-2024. Tencent: Hy3 preview is a text model from Tencent, released on 22 April 2026 under the model id tencent/hy3-preview. Both are served as text-in, text-out chat models on api14, so neither accepts images or audio on input, and neither produces anything other than text on output.
The clearest split between them is context and reasoning. Hy3 preview carries a context window of 262144 tokens, twice the 128000 tokens Command R offers, so it is the one to reach for when a single request has to hold very long documents, large codebases or extended conversation history. Hy3 preview also exposes reasoning controls (include_reasoning, reasoning and reasoning_effort), which Command R does not, making it the only one of the two that can be asked to think through a problem before answering. Command R goes the other way on structure: it is the only one of the pair that supports structured outputs and response_format, so when you need the reply to conform to a JSON schema or a fixed shape, it is the natural choice. Their parameter sets otherwise diverge in small ways. Command R also accepts frequency_penalty, presence_penalty, stop and top_k, while Hy3 preview does not list those. Both accept temperature, top_p, max_tokens, seed, tools and tool_choice, so function calling is available on either side.
For day to day work, pick based on what the task actually needs. If you are doing extraction, classification, form filling or anything where the response must land in a strict schema, Command R is the better fit because structured outputs are supported directly rather than coaxed out through prompting. If you are summarizing or querying a very large body of text, or you want the model to work through multi step logic, arithmetic or planning before it commits to an answer, Hy3 preview is the stronger option given its larger context and its reasoning parameters. Tool use is a tie, so agent style workloads that only need function calling can go either way. On price, Command R is the cheaper of the two on input, while output pricing is about the same for both, so cost only tilts the decision when your workload is input heavy.
Calling either model through api14 is the same operation. The endpoint is OpenAI compatible, so you point your existing OpenAI client or SDK at the api14 base URL, send your usual chat completions request, and set the model field to the id you want. Switching between the two means changing that one string, from cohere/command-r-08-2024 to tencent/hy3-preview or back. Your messages array, your system prompt and your streaming setup stay as they are. The only adjustments worth making are the ones the models themselves require: drop response_format and the penalty parameters when you move to Hy3 preview, and drop reasoning_effort and the other reasoning fields when you move to Command R, since each model will reject parameters it does not list.
Switch with one string
Both models use the same endpoint and request shape. Change the model value from "cohere/command-r-08-2024" to "tencent/hy3-preview" and prompts, streaming and tools keep working.
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api14.com/api/v1",
apiKey: process.env.API14_API_KEY,
});
const result = await client.chat.completions.create({
"model": "tencent/hy3-preview",
"messages": [
{
"role": "user",
"content": "Explain vector databases in two sentences."
}
]
});
console.log(result.choices[0].message.content);
More comparisons
- LiquidAI: LFM3-1.7B vs Ricursive: Ripple 65B Preview
- LiquidAI: LFM3-1.7B vs Crouton Alpha
- LiquidAI: LFM3-1.7B vs Tencent: Hy4
- LiquidAI: LFM3-1.7B vs Z.ai: GLM 5.3 FlashX Legal
- Ricursive: Ripple 65B Preview vs Crouton Alpha
- Ricursive: Ripple 65B Preview vs Tencent: Hy4
- Ricursive: Ripple 65B Preview vs Z.ai: GLM 5.3 FlashX Legal
- Crouton Alpha vs Tencent: Hy4