Home/Models
Models/tencent

Hy3

Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort: a direct no-think mode by default, plus low and high chain-of-thought modes for complex math, coding, and multi-step problems. With a 256K context window, Hy3 targets long-horizon tasks, including improved coreference resolution, multi-turn constraint tracking, and stable tool-calling that generalizes across agent scaffoldings.

Tencent positions it as a reliable, cost-effective option across coding, document processing, financial analysis, game development, and frontend design, with a strong emphasis on grounded, anti-hallucination behavior that answers when grounded and flags when evidence is missing rather than fabricating.

tencent/hy3
Organization
tencent
Family
Hy
Providers
20
Context
256,000
Output limit
128,000
Knowledge
—
Release
2026-07-06
Updated
2026-07-06
Weights
Open
Input
text
Output
text
Capabilities
Tools, Reasoning, Temperature

Providers

Compare every cataloged offer for this model, including its provider, route, token limits, price, and capabilities.

30 offers · USD per million tokens
ProviderModel IDContextOutputInput / output · 1MReasoningToolsStructuredDetails
AIHubMix
hy3
256,000128,000$0.1562 / $0.6248YesYesYes
View
CrossModel
tencent/hy3
262,144131,072$0.16 / $0.64YesYesYes
View
Deep Infra
tencent/Hy3
262,144128,000$0.13 / $0.53YesYesYes
View
DevPass (LLM Gateway)
hy3
262,144128,000$0.132 / $0.528YesYes—
View
Eden AI
deepinfra/tencent/Hy3
262,144128,000$0.13 / $0.53YesYesYes
View
Eden AI
tencent/hy3
256,000128,000$0.132 / $0.528YesNoNo
View
Hugging Face
tencent/Hy3
262,144128,000$0.14 / $0.58YesYesYes
View
Jalapeno Cloud
Hy3
202,752128,000$0.14 / $0.58YesYesYes
View
Kenari
hy3
256,000128,000$0.00 / $0.00YesYes—
View
Kenari
hy3:free
256,000128,000$0.00 / $0.00YesYes—
View
Kilo Gateway
tencent/hy3
262,144128,000$0.13 / $0.53YesYesYes
View
LLM Gateway
novita/hy3
262,144262,144$0.14 / $0.58YesYesYes
View
LLM Gateway
deepinfra/hy3
262,144131,072$0.14 / $0.58YesYesYes
View
LLM Gateway
tencent/hy3
262,144131,072$0.132 / $0.528YesYesYes
View
NanoGPT
tencent/hy3
262,144128,000$0.066 / $0.26YesYesYes
View
OpenCode Go
hy3
256,000128,000$0.14 / $0.58YesYes—
View
OpenCode Zen
hy3-free
190,00064,000$0.00 / $0.00YesYesYes
View
AtlasCloudvia OpenRouter
tencent/hy3
fp8		262,144131,072$0.20 / $0.80YesYesYes
View
DeepInfravia OpenRouter
tencent/hy3
fp4		262,144131,072$0.13 / $0.53YesYesYes
View
GMICloudvia OpenRouter
tencent/hy3
bf16		262,144235,929$0.14 / $0.58YesYesNo
View
Novitavia OpenRouter
tencent/hy3
unknown		262,144235,929$0.14 / $0.58YesYesYes
View
Phalavia OpenRouter
tencent/hy3
unknown		262,144235,929$0.15 / $0.64YesYesYes
View
Tencentvia OpenRouter
tencent/hy3
fp8		262,144128,000$0.132 / $0.528YesYesNo
View
OrcaRouter
tencent/hy3
256,000128,000$0.18 / $0.59YesYes—
View
OrcaRouter
tencent/hy3-free
256,000128,000$0.00 / $0.00YesYes—
View
Requesty
hy3
262,144262,144$0.14 / $0.58YesYesYes
View
SiliconFlow
tencent/Hy3
262,144262,144$0.132 / $0.528YesYes—
View
Tencent TokenHub
hy3
256,000128,000$0.00 / $0.00YesYes—
View
Tencent Token Plan
hy3
256,000128,000$0.00 / $0.00YesYes—
View
Vercel AI Gateway
tencent/hy3
262,144262,144$0.14 / $0.58YesYes—
View

Prices and limits apply to each offer. A dash means the value is not provided. OpenRouter offers show the hosting provider and routing channel separately.

Pricing

Compare token rates and cache charges across available routes and hosting configurations.

USD per million tokens
Input from$0.00Per 1M input tokens
Output from$0.00Per 1M output tokens
Available offers30Across 20 providers

Lowest listed input and output prices across the offers above; they may belong to different providers. These are base rates, before conditional pricing, discounts or additional fees.

Endpoint price comparisonLatest observed snapshot · 6 lowest-priced configurations
InputOutput
DeepInfradeepinfra/fp4
$0.13$0.53
Tencenttencent/fp8
$0.132$0.528
GMICloudgmicloud/bf16
$0.14$0.58
Novitanovita
$0.14$0.58
Phalaphala
$0.15$0.64
AtlasCloudatlas-cloud/fp8
$0.20$0.80
tencent/hy3Observed
Input
$0.132
Output
$0.528
Cache read
$0.033
Additional fees and pricing conditions

overrides · USD per token for token prices; time windows use UTC.

[
  {
    "prompt": "0.000000132",
    "utc_end": 1600,
    "utc_start": 0,
    "completion": "0.000000528",
    "input_cache_read": "0.000000033"
  },
  {
    "prompt": "0.0000000825",
    "utc_end": 0,
    "utc_start": 1600,
    "completion": "0.00000033",
    "input_cache_read": "0.000000020625"
  }
]
Hosting prices through OpenRouter All endpoints and cache rates
EndpointInput / 1MOutput / 1MCache read / 1MCache write / 1M
AtlasCloudatlas-cloud/fp8$0.20$0.80$0.05—
DeepInfradeepinfra/fp4$0.13$0.53$0.033—
GMICloudgmicloud/bf16$0.14$0.58$0.035—
Novitanovita$0.14$0.58$0.035—
Phalaphala$0.15$0.64$0.04—
Tencenttencent/fp8$0.132$0.528$0.033—

Each row is a hosting configuration. Open an offer in Providers for its full pricing conditions and additional fees. Endpoint observations may differ from the routing catalog's base price.

Performance

Review reported latency and output-throughput percentiles from the latest observed rolling window.

Latest 30-minute observation

Time to first token · milliseconds

Latency distributionRolling 30-minute snapshot · 6 endpoints
P50P75P90P99
AtlasCloudatlas-cloud/fp8
2,076–4,915 ms
DeepInfradeepinfra/fp4
836–6,852.06 ms
GMICloudgmicloud/bf16
2,462–23,111.01 ms
Novitanovita
3,677–31,508.21 ms
Phalaphala
2,157–3,673.3 ms
Tencenttencent/fp8
3,579–22,879.8 ms
All latency percentiles 6 endpoints
EndpointP50P75P90P99
AtlasCloudatlas-cloud/fp82,076 ms2,393.5 ms2,566 ms4,915 ms
DeepInfradeepinfra/fp4836 ms1,262.25 ms1,691 ms6,852.06 ms
GMICloudgmicloud/bf162,462 ms3,126.25 ms3,835.4 ms23,111.01 ms
Novitanovita3,677 ms5,090.5 ms6,163.8 ms31,508.21 ms
Phalaphala2,157 ms2,604 ms2,915 ms3,673.3 ms
Tencenttencent/fp83,579 ms4,579.25 ms6,359.8 ms22,879.8 ms

Output throughput · tokens per second

Throughput distributionRolling 30-minute snapshot · 6 endpoints
P50P75P90P99
AtlasCloudatlas-cloud/fp8
88–126.44 t/s
DeepInfradeepinfra/fp4
110–202.14 t/s
GMICloudgmicloud/bf16
93–137.98 t/s
Novitanovita
65–162 t/s
Phalaphala
85–112.45 t/s
Tencenttencent/fp8
63–158 t/s
All throughput percentiles 6 endpoints
EndpointP50P75P90P99
AtlasCloudatlas-cloud/fp888 t/s107 t/s117.8 t/s126.44 t/s
DeepInfradeepinfra/fp4110 t/s142 t/s165.4 t/s202.14 t/s
GMICloudgmicloud/bf1693 t/s110 t/s118 t/s137.98 t/s
Novitanovita65 t/s91 t/s108 t/s162 t/s
Phalaphala85 t/s97.25 t/s105.5 t/s112.45 t/s
Tencenttencent/fp863 t/s87 t/s110.9 t/s158 t/s

Percentiles describe the observed request distribution. These are snapshots of a rolling window, not a historical time series.

Uptime

Compare successful request rates across the latest 5-minute, 30-minute, and 24-hour observations.

Successful requests by window
Availability by endpointCurrent overlapping windows · 6 endpoints
Endpoint5 minutes30 minutes24 hours
AtlasCloudatlas-cloud/fp8
DeepInfradeepinfra/fp4
GMICloudgmicloud/bf16
Novitanovita
Phalaphala
Tencenttencent/fp8

Reported by OpenRouter at the time of collection; rate-limited requests are excluded. A dash means no measurement was reported. This is not a live availability check.

Benchmarks

Review attributed evaluation results on each source's original scale.

External evaluations
tencent/hy3Observed
Design ArenaCategoryEloWin rateRank
models3d1,19143.9%50
modelscodecategories1,18341.1%60
modelsdataviz1,13436.1%84
modelsgamedev1,14938.6%71
modelsuicomponent1,17040.2%67
modelswebsite1,18941.3%65

Sources: Artificial Analysis and Design Arena, via OpenRouter. Scores retain their original scales and are not OpenAI Suite ratings.

Detailed evaluations

EvaluationConfigurationSourceResultsDetails
models · websiteHy3 Lowtencent/hy3-20260706Design Arena1,189 Elo
Details
models · gamedevHy3 Lowtencent/hy3-20260706Design Arena1,149 Elo
Details
models · uicomponentHy3 Lowtencent/hy3-20260706Design Arena1,170 Elo
Details
models · datavizHy3 Lowtencent/hy3-20260706Design Arena1,134 Elo
Details
models · codecategoriesHy3 Lowtencent/hy3-20260706Design Arena1,183 Elo
Details
models · 3dHy3 Lowtencent/hy3-20260706Design Arena1,191 Elo
Details

Sources: Artificial Analysis, Design Arena and OpenRouter. Evaluation configurations and scales differ; results are not interchangeable. Updated Oct 4, 2026.

Apps & session costs

Compare observed median session costs for applications using this model, grouped by conversation length.

30-day sample · through Sep 27, 2026
ApplicationModel routeTurnsMedian cost / sessionDetails
Hermes Agenttencent/hy3-202607062–9 turns
$0.006984
Details
Hermes Agenttencent/hy3-2026070650+ turns
$0.280488
Details
Hermes Agenttencent/hy3-2026070610–49 turns
$0.047433
Details
Hermes Agenttencent/hy3-202607061 turn
$0.001629
Details

Source: OpenRouter, as of Oct 2, 2026. Published under CC BY 4.0. These are observed costs per session, not token prices or a forecast for your workload. Apps are compared separately.

Activity

Daily prompt and completion tokens reported for this model's routes in OpenRouter's top 50. Missing days have no published value.

Last 90 complete UTC days
Daily token usage58,651,078,356,054 reported tokens · 75 days with observations
Jul 6, 2026Oct 3, 2026
Daily observations · 75 days
Date (UTC)Reported tokensReported routes
Oct 3, 2026188,077,314,4861
Oct 2, 2026207,199,060,6071
Oct 1, 2026223,460,678,3611
Sep 30, 2026395,954,245,4791
Sep 29, 2026437,823,584,2911
Sep 28, 2026430,830,913,9341
Sep 27, 2026262,624,758,6171
Sep 26, 2026232,398,920,4421
Sep 25, 2026217,363,957,6481
Sep 24, 2026415,575,685,9141
Sep 23, 2026456,483,783,6931
Sep 22, 2026482,617,399,2111
Sep 21, 2026501,869,218,0161
Sep 20, 2026487,810,242,2961
Sep 19, 2026484,255,983,4531
Sep 18, 2026728,437,121,8131
Sep 17, 2026766,024,944,2191
Sep 16, 2026756,338,132,3511
Sep 15, 2026782,378,301,8551
Sep 14, 2026775,665,854,9521
Sep 13, 2026458,068,532,8511
Sep 12, 2026473,250,473,6991
Sep 11, 2026678,292,982,1621
Sep 10, 2026547,218,166,1061
Sep 9, 2026570,557,417,8891
Sep 8, 2026575,153,523,3081
Sep 7, 2026494,937,455,8071
Sep 6, 2026301,392,865,3181
Sep 5, 2026343,412,337,4731
Sep 4, 2026509,687,952,8861
Sep 3, 2026767,864,805,7061
Sep 2, 2026730,117,549,8571
Sep 1, 2026515,136,778,2131
Aug 31, 2026824,593,260,2981
Aug 30, 2026748,368,879,5691
Aug 29, 2026699,094,605,9351
Aug 28, 2026963,010,163,2921
Aug 27, 20261,108,337,370,6931
Aug 26, 20261,035,208,001,4101
Aug 25, 20261,030,807,763,9531
Aug 24, 20261,071,620,753,5961
Aug 23, 2026707,113,475,9651
Aug 22, 2026753,953,351,5201
Aug 21, 20261,122,505,219,6161
Aug 20, 20261,254,528,836,0091
Aug 19, 20261,216,315,602,2661
Aug 18, 20261,634,318,224,0141
Aug 17, 20261,517,431,295,8721
Aug 16, 20261,064,840,646,0491
Aug 15, 20261,117,668,844,0801
Aug 14, 20261,418,783,594,6031
Aug 13, 20261,497,101,053,0221
Aug 12, 20261,577,744,350,2001
Aug 11, 20261,510,734,193,3951
Aug 10, 20261,766,439,374,9391
Aug 9, 20261,551,112,637,8641
Aug 8, 20261,466,191,716,6061
Aug 7, 20261,275,944,731,3711
Aug 6, 2026949,493,504,4201
Aug 5, 2026964,850,842,8581
Aug 4, 2026962,249,536,9201
Aug 3, 2026880,285,616,8091
Aug 2, 2026631,248,510,1321
Aug 1, 2026580,174,036,7081
Jul 31, 2026724,961,260,8491
Jul 30, 2026612,482,883,7891
Jul 29, 2026613,750,145,2911
Jul 28, 2026842,114,840,4031
Jul 27, 2026819,849,057,4541
Jul 26, 2026597,279,135,4081
Jul 25, 2026590,032,165,7821
Jul 24, 2026717,183,543,7231
Jul 23, 2026723,167,189,7031
Jul 22, 2026713,187,597,7261
Jul 21, 2026596,719,599,0291

Source: OpenRouter (openrouter.ai/rankings), as of Oct 4, 2026. CC BY 4.0. Totals include only individually published routes. Tokenizers differ by provider; the aggregated “other” category is never assigned to a model.

Technical details & API

Explore specifications, supported parameters, reasoning controls, and API identifiers for each OpenRouter route.

Current route specifications
tencent/hy3Observed
Context
262,144
Maximum output
128,000
Tokenizer
Other
Input
text
Output
text
Moderation
No
Added to OpenRouter
2026-07-06
Base URLhttps://openrouter.ai/api/v1
Model IDtencent/hy3
Version IDtencent/hy3-20260706
Weights IDtencent/Hy3

Reasoning

Required
No
Enabled by default
Yes
Default effort
high
Supported efforts
high, low, none

Supported parameters

  • frequency_penalty
  • include_reasoning
  • logit_bias
  • max_completion_tokens
  • max_tokens
  • min_p
  • presence_penalty
  • reasoning
  • reasoning_effort
  • repetition_penalty
  • response_format
  • seed
  • stop
  • structured_outputs
  • temperature
  • tool_choice
  • tools
  • top_k
  • top_p

Default parameters

top p
1
temperature
0.9

Frequently asked questions

Answers based on the model metadata and serving offers currently in the catalog.

How much does Hy3 cost?

Listed rates start at $0.00 per million input tokens and $0.00 per million output tokens. Prices vary by provider and configuration; see Pricing for details.

Which providers offer Hy3?

There are 30 cataloged offers across 20 providers. Compare model IDs, prices and limits in Providers.

What is the context window?

The cataloged model context is 256,000 tokens. Each serving endpoint may apply a different limit.

Does it support tools and structured output?

Tool calling: Yes. Structured output: Not reported. Support can vary by endpoint.

Resources

Hugging Face ↗
Catalog history · 2 recorded revisions

Metadata observations since this model was first cataloged. These are not model release versions.

  • · 07993e7081d9
  • · c136514cd0ef