Skip to content
AI Atlas
ModelActiveOpen weightsIdentity probable

Hy3

Tencenthuggingface.co/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:...

Open in Graph
data quality72

Updated 3 h ago · first seen 11 Sept 2026

model_01M294WVW4C092JT6V607VWDPF

Overview

Identity

Canonical model
Yesidentity confidence: mediumOne row per real model release. Artifacts (checkpoints, quantisations, conversions) and folded evaluation variants point here.
Official checkpoints
official_checkpoints = hf_repo identifiers carried by the model itself; artifacts are separate entities pointing here through canonical_id.
Artifacts
None recordedSeparate entities (checkpoint · quantization · conversion · packaging) pointing to this model through canonical_id.
Provider deployments
1
API aliases
hy3hy3-non-reasoningtencent/hy3Identifiers under which providers and evaluators refer to this model.
Folded evaluation variants
1Effort / thinking variants (…-high, …-non-reasoning) are result configurations of this model, not separate models. Their old URLs redirect here.

Openness

Open weights— weights downloadable under Apache-2.0; commercial use allowed; redistribution allowed; derivatives allowed; 4 dimensions unknown.

Weights downloadable under a permissive or Creative Commons licence allowing commercial use; code or data may be missing.

  • Weights

    Yes

  • Inference code

    —

  • Training code

    —

  • Training data

    —

  • Dataset

    —

  • Commercial use

    Yes

  • Redistribution

    Yes

  • Derivatives

    Yes

Licence: Apache License 2.0 (permissive · SPDX Apache-2.0 · stated as “apache-2.0”)

dimensions marked null are unknown, not false

Key facts

Release date

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 7 h agomedium

Status

Source:Artificial AnalysisT2observed 11 h agomedium

Model card
huggingface.co/tencent/Hy3

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium

Openrouter id

Source:OpenRouter public model & pricing listingT2observed 13 h agomedium

Architecture

Architecture

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Model type

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Parameters

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium

Weights dtype

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

File size

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Library name

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 12 h agomedium

Pipeline tag

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 13 h agomedium

Hugging Face repo

Source:OpenRouter public model & pricing listingT2observed 13 h agomedium

Capabilities

Modalities

Modalities
text
Input
text
Output
text

Capabilities

  • Tool calling

    Yes

    OpenRouter public model & pricing listing · T2

  • Structured output

    Yes

    OpenRouter public model & pricing listing · T2

  • Reasoning

    Yes

    Artificial Analysis · T2

  • Vision

    Unavailable

  • Audio

    Unavailable

  • Fine-tuning available

    Unavailable

Context window

Source:OpenRouter public model & pricing listingT2observed 10 h agomedium

Max output

Source:OpenRouter public model & pricing listingT2observed 13 h agomedium

Benchmarks30

Compare with another model →

Comparable same task and conditions · Partially comparable same task, conditions differ (effort, temperature, judge) · Not comparable different variant or metric

Benchmark results grouped by comparability group
Benchmark · groupBest scoreTrustConfigurationResultsvs leaderEvaluatedSource
Terminal-Benchagentic · accuracy · variant=hard · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisvarianthardreasoningonconditions differ across rows → partially comparable3−31.8 ptvs gpt-5.6-solobs. 12 Sept 2026artificialanalysis.aiT2
Terminal-Benchagentic · accuracy · variant=v2.1 · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisvariantv2.1reasoningonconditions differ across rows → partially comparable2non-primary groupobs. 12 Sept 2026artificialanalysis.aiT2
Terminal-Benchagentic · accuracy · variant=v4.0 · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisvariantv4.0reasoningonconditions differ across rows → partially comparable2non-primary groupobs. 12 Sept 2026artificialanalysis.aiT2
τ²-benchagentic · pass^1 · variant=Telecom · evaluator=Artificial AnalysisIndependentevaluatorArtificial AnalysisvariantTelecomreasoningonconditions differ across rows → partially comparable2non-primary groupobs. 12 Sept 2026artificialanalysis.aiT2
τ²-benchagentic · pass^1 · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisreasoningoffconditions differ across rows → partially comparable1−31.6 ptvs Z.ai GLM 5.2obs. 11 Sept 2026artificialanalysis.aiT2
SciCodecoding · accuracy · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisreasoningonconditions differ across rows → partially comparable2−14.5 ptvs Claude Fable 5.1obs. 12 Sept 2026artificialanalysis.aiT2
Artificial Analysis Intelligence Indexcomposite · indexIndependentreasoningonversion4.3conditions differ across rows → partially comparable5−27.6vs Claude Fable 5.1obs. 12 Sept 2026artificialanalysis.aiT2
IFBenchinstruction-following · accuracy · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisreasoningonconditions differ across rows → partially comparable3−20.2 ptvs Grok 4.3obs. 12 Sept 2026artificialanalysis.aiT2
Humanity's Last Examknowledge · accuracy · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisreasoningonconditions differ across rows → partially comparable5−25.7 ptvs Claude Fable 5.1obs. 12 Sept 2026artificialanalysis.aiT2
GPQA Diamondreasoning · accuracy · variant=Diamond · evaluator=Artificial AnalysisIndependentevaluatorArtificial AnalysisvariantDiamondreasoningonconditions differ across rows → partially comparable3non-primary groupobs. 12 Sept 2026artificialanalysis.aiT2
GPQA Diamondreasoning · accuracy · variant=GPQA Diamond · evaluator=Artificial AnalysisIndependentevaluatorArtificial AnalysisvariantGPQA Diamond2−6.56 ptvs gpt-6-astraobs. 11 Sept 2026artificialanalysis.aiT2

Current rows only, grouped by benchmark → canonical metric → comparability group (task configuration). Effort variants folded into this model appear as rows of the same group. 30 current rows in total. “vs leader” compares with the current leader of the benchmark's primary group only; other groups are not directly comparable. Comparability rules →

Providers & Pricing1

All offers in the price terminal →
Provider deployments, cheapest output first
ProviderContextInput / 1MCached inOutput / 1MStatusObservedSource
OpenRoutercheapest outputtencent/hy3262.1Kout 128K$0.033active20 min agosince 12 Sept 2026openrouter.aiT2

USD per 1M tokens as published by each provider; native units (per-request fees, flex/priority tiers) are kept verbatim. Rows are append-only — every price change is kept in the history below. Cost of a workload →

Price history

Step lines per provider; amber markers are recorded changes. Click a marker or a row for the evidence behind that price.

Output price · USD / 1M tokens 1 provider

Output price history of Hy3$0$0.20$0.40$0.60$0.80Sept 26Sept 26Sept 26Sept 26OpenRouter: first observed → $0.33 · 11 Sept 2026OpenRouter: $0.33 → $0.528 · 12 Sept 2026
  • OpenRouter
  • OpenRouter$0.33 → $0.52812 Sept 2026
  • OpenRouterfirst observed $0.3311 Sept 2026

Input price · USD / 1M tokens 1 provider

Input price history of Hy3$0$0.05$0.10$0.15$0.20Sept 26Sept 26Sept 26Sept 26OpenRouter: first observed → $0.083 · 11 Sept 2026OpenRouter: $0.083 → $0.132 · 12 Sept 2026
  • OpenRouter
  • OpenRouter$0.083 → $0.13212 Sept 2026
  • OpenRouterfirst observed $0.08311 Sept 2026

Hardware fit37

Estimated

7 of 37 device × quantization combinations fit.

Run locally: your machine →
Estimated hardware fit
HardwareQuantizationDevice memoryEst. memoryFits
Apple M3 Ultra4bit—172.3 GB est.Yes
Mac Studio (Apple M5 Ultra)4bit—172.3 GB est.Yes
Apple M2 Ultra4bit—172.3 GB est.Yes
NVIDIA B2004bit180 GB172.3 GB est.Yes
AMD Instinct MI300X4bit192 GB172.3 GB est.Yes
AMD Instinct MI325X4bit256 GB172.3 GB est.Yes
NVIDIA DGX B2004bit1,440 GB172.3 GB est.Yes
Apple M1 Ultra4bit—172.3 GB est.No
Apple M3 Max4bit—172.3 GB est.No
Apple M4 Max4bit—172.3 GB est.No
Mac Studio (Apple M5 Max)4bit—172.3 GB est.No
MacBook Pro (Apple M5 Max)4bit—172.3 GB est.No
Apple M2 Max4bit—172.3 GB est.No
Apple M1 Max4bit—172.3 GB est.No
Apple M4 Pro4bit—172.3 GB est.No
Mac mini (Apple M5 Pro)4bit—172.3 GB est.No
MacBook Pro (Apple M5 Pro)4bit—172.3 GB est.No
Apple M3 Pro4bit—172.3 GB est.No
Apple M1 Pro4bit—172.3 GB est.No
Apple M2 Pro4bit—172.3 GB est.No
Apple M44bit—172.3 GB est.No
iMac (Apple M4)4bit—172.3 GB est.No
Mac mini (Apple M6)4bit—172.3 GB est.No
MacBook Air (Apple M5)4bit—172.3 GB est.No
MacBook Pro (Apple M5)4bit—172.3 GB est.No
Apple M24bit—172.3 GB est.No
Apple M34bit—172.3 GB est.No
Apple M14bit—172.3 GB est.No
NVIDIA GeForce RTX 30904bit24 GB172.3 GB est.No
NVIDIA GeForce RTX 40904bit24 GB172.3 GB est.No
NVIDIA GeForce RTX 50904bit32 GB172.3 GB est.No
NVIDIA A100 80GB4bit80 GB172.3 GB est.No
NVIDIA H100 SXM4bit80 GB172.3 GB est.No
NVIDIA H100 NVL4bit94 GB172.3 GB est.No
NVIDIA DGX Spark4bit128 GB172.3 GB est.No
NVIDIA H2004bit141 GB172.3 GB est.No
NVIDIA H200 NVL4bit141 GB172.3 GB est.No
Assumptions (7)
  • Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
  • bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
  • KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
  • A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
  • Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
  • Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
  • Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.

Versions & Artifacts0

Version history

Context window2 changes

11 Sept 2026→11 Sept 2026→12 Sept 2026current

Licensefirst observation only

11 Sept 2026current

Max outputfirst observation only

11 Sept 2026current

Opennessfirst observation only

11 Sept 2026current

Parametersfirst observation only

11 Sept 2026current

Statusfirst observation only

11 Sept 2026current

Each hop is a claim: click a value for its source, tier and observation time. Nothing is overwritten — a new observation closes the previous claim.

Artifacts 0

No artifact (checkpoint, quantisation, conversion or packaging) points to this model yet.

Timeline23

Full timeline →
  • Benchmark resultModelHy3Tencent

    Hy3 scores 34.09% on Terminal-Bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 92.69% on τ²-bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 63.13% on IFBench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 27.76% on Humanity's Last Exam

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 86.67% on GPQA Diamond

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 22.74 on Artificial Analysis Intelligence Index

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 64.42% on Terminal-Bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 0.51% on Terminal-Bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 48.61% on SciCode

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 33.46% on Humanity's Last Exam

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 89.7% on GPQA Diamond

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 25.77 on Artificial Analysis Intelligence Index

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 31.82% on Terminal-Bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 67.54% on τ²-bench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 47.96% on IFBench

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 6.95% on Humanity's Last Exam

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 73.23% on GPQA Diamond

    artificial_analysis
    artificialanalysis.ai
  • Benchmark resultModelHy3Tencent

    Hy3 scores 17.04 on Artificial Analysis Intelligence Index

    artificial_analysis
    artificialanalysis.ai
  • Property changedModelHy3Tencent

    Hy3: reasoning changed from false to true

    ReasoningNo→Yesartificial_analysis
    artificialanalysis.ai
  • Price changedModelHy3Tencent

    OpenRouter changed pricing for Hy3: $0.0825 in / $0.33 out per 1M tokens → $0.132 in / $0.528 out per 1M tokens

    $0.083 in / $0.33 out→$0.132 in / $0.528 outopenrouter
    openrouter.ai
  • Context window changedModelHy3Tencent

    Hy3: context length changed from 256000 to 262144

    Context window256K tokens→262.1K tokensopenrouter
    openrouter.ai
  • Release date changedModelHy3Tencent

    Hy3: release date changed from 2026-07-02 to 2026-07-06

    Release date2 Jul 2026→6 Jul 2026openrouter
    openrouter.ai
  • Release date changedModelHy3Tencent

    Hy3: release date changed from 2026-07-06 to 2026-07-02

    Release date6 Jul 2026→2 Jul 2026huggingface
    huggingface.co

Change history

Temporal, append-only claims: a new observation closes the previous claim instead of overwriting it. Rewind the record with the as-of picker.
1 claims · 1 propertiesShow all properties

Redistribution allowedredistribution_allowed1

Claim history for Redistribution allowed
ValueValid from → toStatusSourceConfidenceExtractor
Yes → currentcurrentAI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2highderived

Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →

Provenance

Attributed facts

41

Source tiers

T241

Freshest observation

3 h ago

Conflicts

None

Source documents 5

Source documents
SourceDocumentTypeTierLast observedSnapshots
OpenRouter public model & pricing listingopenrouter.ai/api/v1/models catalogueT2· Quality secondary20 min ago10
Artificial Analysisartificialanalysis.ai/leaderboards/models leaderboardT2· Quality secondary3 h ago2
Hugging Face Hub (public pages, model cards, papers)huggingface.co/models?author=tencent&p=0&sort=downloads listingT2· Quality secondary3 h ago7
Hugging Face Hub (public pages, model cards, papers)huggingface.co/tencent/Hy3/raw/main/README.md model_cardT2· Quality secondary6 h ago1
Hugging Face Hub (public pages, model cards, papers)huggingface.co/tencent/Hy3 model_pageT2· Quality secondary6 h ago5

Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.

Data quality (72/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →