Hy3-preview
Hy3-preview is capable in long context and coding; and behind the leaders in reasoning, factuality, agentic tasks, and instruction following. Too few results yet to rate safety, math, multimodal tasks, or multilingual tasks.
Capability profile
Bars show the model's median result as a share of the leading model's, per capability. Select a row to see its results.
Too few results yet to rate Safety, Math, Multimodal or Multilingual.
Price
$0.14input$0.56outputper million tokens
From Artificial Analysis · 3 providers tracked · All prices
Evidence
48results on33benchmarks
- 1 independently verified
- 30 aggregator
- 17 vendor-reported
From 3 sources · latest Oct 8, 2026 · How verification works
API features
Tool callingReasoning
As listed by OpenRouter
Hy3-preview benchmark results
48 results on 33 benchmarks, grouped by capability. Each bar is the result as a share of the capability leader's; every result links to its source.
13.2% behind the leader1 of 3 ranked benchmarks measured
- 79.00Oct 8, 2026
23.3% behind the leader7 of 10 ranked benchmarks measured
- 78.00Oct 7, 2026
- 75.80Oct 7, 2026
- SciCode48.61Oct 8, 2026aa_scicode
- 1508.15May 22, 2026
- Terminal-Bench 2.164.42Oct 8, 2026terminalbenchV21
- 57.90Oct 7, 2026
- Terminal-Bench Hard31.82Oct 8, 2026aa_terminalbench_hard
Show 2 more coding resultsHide 2 coding results
- SciCode39.35Sep 4, 2026aa_scicode
- 71.70Oct 7, 2026
25.1% behind the leader3 of 6 ranked benchmarks measured
- GPQA Diamond89.70Oct 8, 2026gpqa
- Humanity's Last Exam33.46Oct 8, 2026aa_hle
- 4.86Oct 8, 2026
Show 4 more reasoning resultsHide 4 reasoning results
- 0.29Oct 8, 2026
- GPQA Diamond73.23Oct 8, 2026gpqa
- GPQA Diamond90.40Oct 7, 2026GPQA
- Humanity's Last Exam6.95Oct 8, 2026aa_hle
27.1% behind the leader2 of 4 ranked benchmarks measured
- AA-Omniscience · Accuracy31.97Oct 8, 2026omniscienceAccuracy
- AA-Omniscience · Non-hallucination25.87Oct 8, 2026omniscienceNonHallucination
Show 2 more factuality resultsHide 2 factuality results
- AA-Omniscience · Accuracy23.18Oct 8, 2026omniscienceAccuracy
- AA-Omniscience · Non-hallucination24.32Oct 8, 2026omniscienceNonHallucination
28.3% behind the leader5 of 7 ranked benchmarks measured
- 84.20Oct 7, 2026
- 79.10Oct 7, 2026
- τ-Bench V3 · Banking22.89Oct 8, 2026tauBanking
- 28.34Oct 8, 2026
- Terminal-Bench 4.00.51Oct 8, 2026
Show 2 more agentic resultsHide 2 agentic results
- AA ApexAgents25.60Oct 7, 2026APEX-Agents
- 36.45Jun 15, 2026
40.1% behind the leader1 of 3 ranked benchmarks measured
- IFBench47.96Oct 8, 2026aa_ifbench
0 of 5 ranked benchmarks measured
- 90.00Oct 7, 2026
More results
Benchmarks outside the capability baskets. They are not ranked against a leader.
- τ²-Bench Telecom (AA run)67.54Oct 8, 2026aa_tau2
- AA Intelligence25.30Oct 8, 2026aa_intelligence_index
- -18.47Oct 8, 2026
- -34.95Oct 8, 2026
- AA Intelligence17.04Oct 8, 2026aa_intelligence_index
- Artificial Analysis Coding Index58.80Sep 9, 2026aa_coding_index
Show 10 more resultsHide 10 results
- 25.64Sep 9, 2026
- 46.70Jun 18, 2026
- 34.33Jun 18, 2026
- Claw Eval (pass@3)68.50Oct 7, 2026Claw-Eval
- DeepSearchQA (F1)91.00Oct 7, 2026DeepSearchQA
- 28.00Oct 7, 2026
- MathArena Apex (Pass@1)38.70Oct 7, 2026MathArena Apex
- 45.60Oct 7, 2026
- 48.50Oct 7, 2026
- 76.40Oct 7, 2026
Hy3-preview: common questions
Who makes Hy3-preview?
Hy3-preview is made by Tencent.
When was Hy3-preview released?
Hy3-preview was released on Apr 23, 2026, according to Artificial Analysis.
What is Hy3-preview good at?
Hy3-preview is capable in long context and coding; and behind the leaders in reasoning, factuality, agentic tasks, and instruction following. Too few results yet to rate safety, math, multimodal tasks, or multilingual tasks.
How much does Hy3-preview cost?
Hy3-preview costs $0.14 per million input tokens and $0.56 per million output tokens, according to Artificial Analysis. We track its price at 3 providers. At a mix of three input tokens to one output token, it is cheaper than 78% of the 331 priced models we track.
How many benchmarks has Hy3-preview been tested on?
We track 48 results for Hy3-preview on 33 benchmarks from 3 sources, 1 of them independently verified. The latest was recorded on Oct 8, 2026.
Which API features does Hy3-preview support?
OpenRouter lists tool calling and reasoning for Hy3-preview.
About this record
Where Hy3-preview's numbers come from, and every name it appears under.
- Tracked since
- May 2, 2026
- Newest source mention
- Aug 12, 2026
Where the results come from
Verification: 48 scores · 1 independently verified · 30 aggregator-attributed · 17 vendor-reported. How these tiers are assigned
From 3 sources on 3 sites. Artificial Analysis supplies 30 of them; the 1 independently verified result comes from 1 site. Bars are coloured by trust tier.
- artificialanalysis.ai30
- api.llm-stats.com17
- datasets-server.huggingface.co1