hy3
Tencent/🇸🇬 Singapore/chat
Hy3 is Tencent's open-weight (Apache 2.0) MoE model with a 256K context window, optimized for agentic workflows, coding and productivity tasks with strong token efficiency.MoreLess
Which id to call
tencent/hy3This exact deployment on Tencent, with no routing and no failover. Send it as the model field.
Input /1M
$0.13
$0.03 cached
Output /1M
$0.53
4.0x input
Context
262K
131K output
Added
Jul 2026
chat
Capabilities 4/8
Provider rates
What hy3 costs
Provider prices per 1M tokens, updated September 28, 2026.moreless
These are the upstream provider rates. Pay as you go adds 5%, or 0% if you bring your own keys, and there is no per-request fee. Prompt caching and routing change what you pay against these rates, not the rates themselves.
Input /1M
$0.13
Output /1M
$0.53
Cache write /1M
-
Cache read /1M
$0.03
What a workload costs
- 100K input + 10K output
- $0.0185
- 1M input + 100K output
- $0.18
- 10M input + 1M output
- $1.85
At the rates above, before caching. A cache read costs $0.03 per 1M, so repeated context lands under these figures.
model=
Which id to call
One base url, https://router.requesty.ai/v1, and one key for every id here.
tencent/hy3This exact deployment on Tencent, with no routing and no failover. Send it as the model field.
OpenAI compatible
Call it in three lines
Change the base url, use your Requesty key, set the model to any id on the left. Existing OpenAI SDK code needs no other edit.
123456789101112131415from openai import OpenAI client = OpenAI( api_key="YOUR_REQUESTY_API_KEY", base_url="https://router.requesty.ai/v1", ) response = client.chat.completions.create( model="tencent/hy3", messages=[ {"role": "user", "content": "Explain quantum computing in one paragraph."}, ], ) print(response.choices[0].message.content)
Reference
Specs and data terms
What the catalog reports for this deployment, and what the provider does with the traffic.
Released 2026-07-06
Benchmark scores
Benchmarks measure the model, not this endpoint, so they are the same wherever these weights are served.
Artificial Analysis Coding Index: a composite of coding evaluations including LiveCodeBench, SciCode and Terminal-Bench.
Graduate-level physics, chemistry & biology questions designed to resist Googling.
Artificial Analysis Intelligence Index: a composite of multiple evaluations measuring overall model capability.
Scores from official model cards, Artificial Analysis and public leaderboards. They measure specific skills and do not capture every aspect of model quality, so test on your own workload.
Same provider
More from Tencent
Newest first, on the same provider and the same key.
Reference
hy3 questions
How much does hy3 cost?
hy3 is priced at $0.13 per million input tokens and $0.53 per million output tokens when accessed via Requesty. Prompt caching is supported, which can cut effective input cost by up to 90% on repeated context. Those are the upstream provider rates: pay as you go adds 5% on top, or 0% if you bring your own provider keys.
What is the context window of hy3?
hy3 has a context window of 262K tokens, with a maximum output of 131K tokens per response. That's roughly 350 words of input you can fit in a single prompt.
How does hy3 perform on benchmarks?
hy3 scores 89.7% on GPQA Diamond, 58.8% on Coding Index, 48.6% on SciCode. See the full benchmark chart above for results across MMLU Pro, GPQA Diamond, SWE-Bench Verified, HumanEval, MATH, AIME, MMMU, and LiveBench.
What can hy3 do?
hy3 supports tool calling, extended reasoning, prompt caching, structured outputs (JSON schema). You can call it through any OpenAI-compatible client by pointing base_url to Requesty.
How do I use hy3 with the OpenAI SDK?
Install the OpenAI SDK, set base_url to "https://router.requesty.ai/v1", set your API key to your Requesty key, and set the model to "tencent/hy3". The Quickstart above shows Python, JavaScript and cURL snippets.
Can I run hy3 through Requesty?
Yes. hy3 runs through Requesty's OpenAI-compatible API, served from Tencent. You do not host the model yourself: point base_url at Requesty, set the model to "tencent/hy3", and requests are routed to the upstream provider with automatic failover. The same key gives you 600+ other models too.
Call hy3 through one endpoint
One key for this endpoint and 600+ other models. No markup on provider prices, automatic failover, caching built in.