Skip to main content

TencentLanguage model

Hy4 preview

Hy4 preview has no overall index value yet because there are too few independent benchmark results.

PopularityRank 316 of 785

39out of 100-1 vs. last month

based on search, usage and attention

Summary

Hy4 preview targets long-running productivity work. The open-weight Mixture-of-Experts checkpoint activates 49 billion of 770 billion parameters per token and extends the context to 1,048,576 tokens. This checkpoint remains an early release.

Tencent describes Hy4 as an expansion in model size, context length, and training data over Hy3 preview. Tencent publishes vLLM and SGLang recipes. Reasoning defaults to high, while no_think switches to direct responses. Longer-than-necessary reasoning and over-verification are listed as known issues.

See for yourself: one prompt, one attempt

SVG short test

One SVG task with the same fixed settings for every model, without hidden thinking.

Hy4 preview

Hy4 preview: SVG short test

Prompt

Return only one standalone SVG of a pelican riding a bicycle. Use viewBox="0 0 800 600", simple shapes, two clearly visible wheels and a recognizable pelican with a long beak. At most 30 SVG elements and 1800 characters. No Markdown, explanations, comments, scripts, animation, CSS or external resources.

Oct 14, 2026 · $0.0024

Test details
Test run
Oct 14, 2026
Provider
makora/fp8
API model
tencent/hy4-preview
Output
944 tokens
Reasoning
API reports 0 reasoning tokens
Cost of the attempt
$0.0024
Test round
43b8c5c46830
Earlier tests
Hy4 preview: Pelican on a bicycle
A community classic for free-form SVG drawing.

Pelican on a bicycle

Prompt

Generate an SVG of a pelican riding a bicycle

Oct 1, 2026 · 241 s · $0.02

Test details
Run date
Oct 1, 2026
API provider
openrouter
Requested API model
tencent/hy4-preview
Reasoning setting
high
Token limit including reasoning
8,192 tokens
Output tokens
7,009 tokens
Cost of the attempt
$0.0176
Response time
240.6 s
API model reported in response
tencent/hy4-preview
Pinned endpoint
DeepInfra | tencent/hy4-preview-20260827
Temperature
Provider default
Test protocol
community-visual-tests-v1
Technical checks
Passed, not a quality rating

Visually reviewed. This individual test does not establish overall model quality.

Task origin: Simon Willison

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 405 published benchmark results

Page 1 of 34

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

BioMysteryBench: Hard (Show measurement, test conditions, and source)
Source value
39.216
Score
39.216
Metric
accuracy
Unit
%
Benchmark version
1
Category
general
Task
human_difficult
Task label
Hard
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-11-05. Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 2447.966. Standard error: 5.188. Cost per task in USD: 0.597301. Source ID: vals-ai. Evidence capture ID: vals-biomysterybench-2026-11-07-c2992129e13e. Source payload hash: c2992129e13e1455796b51fe354f18fa049ad9dcfe02fd66196230a60aeea9df. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-biomysterybench:1:human_difficult
Comparison methodology
Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
BioMysteryBench: Human-solvable (Show measurement, test conditions, and source)
Source value
76.256
Score
76.256
Metric
accuracy
Unit
%
Benchmark version
1
Category
general
Task
human_solvable
Task label
Human-solvable
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-11-05. Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 1168.673. Standard error: 2.416. Cost per task in USD: 0.306974. Source ID: vals-ai. Evidence capture ID: vals-biomysterybench-2026-11-07-c2992129e13e. Source payload hash: c2992129e13e1455796b51fe354f18fa049ad9dcfe02fd66196230a60aeea9df. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-biomysterybench:1:human_solvable
Comparison methodology
Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
BioMysteryBench: Overall (Show measurement, test conditions, and source)
Source value
69.259
Score
69.259
Metric
accuracy
Unit
%
Benchmark version
1
Category
general
Task
overall
Task label
Overall
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: public. Leaderboard date: 2026-11-05. Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 1410.317. Standard error: 2.593. Cost per task in USD: 0.361814. Source ID: vals-ai. Evidence capture ID: vals-biomysterybench-2026-11-07-c2992129e13e. Source payload hash: c2992129e13e1455796b51fe354f18fa049ad9dcfe02fd66196230a60aeea9df. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-biomysterybench:1:overall
Comparison methodology
Harness: Terminus 2. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration v1.1: CLI (Show measurement, test conditions, and source)
Source value
37.893
Score
37.893
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
coding
Task
cli
Task label
CLI
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-30. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 7813.984. Standard error: 2.57. Cost per task in USD: 4.294849. Source ID: vals-ai. Evidence capture ID: vals-code-migration-v1-1-2026-11-07-4085f8f7e916. Source payload hash: 4085f8f7e916ea8f5f458829ef9b624e3f87c68dc0e79642e03f1c9b679b61c0. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration-v1-1:1.1:cli
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration v1.1: COBOL (Show measurement, test conditions, and source)
Source value
67.778
Score
67.778
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
coding
Task
cobol
Task label
COBOL
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-30. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 4015.106. Standard error: 12.117. Cost per task in USD: 1.435132. Source ID: vals-ai. Evidence capture ID: vals-code-migration-v1-1-2026-11-07-4085f8f7e916. Source payload hash: 4085f8f7e916ea8f5f458829ef9b624e3f87c68dc0e79642e03f1c9b679b61c0. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration-v1-1:1.1:cobol
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration v1.1: Code Quality (Show measurement, test conditions, and source)
Source value
3.408
Score
3.408
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
coding
Task
code_quality
Task label
Code Quality
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-30. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 6864.264. Standard error: 0.058. Cost per task in USD: 3.57992. Source ID: vals-ai. Evidence capture ID: vals-code-migration-v1-1-2026-11-07-4085f8f7e916. Source payload hash: 4085f8f7e916ea8f5f458829ef9b624e3f87c68dc0e79642e03f1c9b679b61c0. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration-v1-1:1.1:code_quality
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration v1.1: Overall (Show measurement, test conditions, and source)
Source value
45.364
Score
45.364
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
coding
Task
overall
Task label
Overall
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-30. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 6864.264. Standard error: 3.59. Cost per task in USD: 3.57992. Source ID: vals-ai. Evidence capture ID: vals-code-migration-v1-1-2026-11-07-4085f8f7e916. Source payload hash: 4085f8f7e916ea8f5f458829ef9b624e3f87c68dc0e79642e03f1c9b679b61c0. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration-v1-1:1.1:overall
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration: CLI (Show measurement, test conditions, and source)
Source value
39.911
Score
39.911
Metric
accuracy
Unit
%
Benchmark version
1
Category
coding
Task
cli
Task label
CLI
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-29. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 9341.195. Standard error: 2.552. Cost per task in USD: 4.20179. Source ID: vals-ai. Evidence capture ID: vals-code-migration-2026-11-07-23c368426e3e. Source payload hash: 23c368426e3e145dbba17219be86d157ba6ebf3b5c25c8f7ba2d5723986fb9b1. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration:1:cli
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration: COBOL (Show measurement, test conditions, and source)
Source value
70
Score
70
Metric
accuracy
Unit
%
Benchmark version
1
Category
coding
Task
cobol
Task label
COBOL
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-29. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 3955.804. Standard error: 15.275. Cost per task in USD: 1.015163. Source ID: vals-ai. Evidence capture ID: vals-code-migration-2026-11-07-23c368426e3e. Source payload hash: 23c368426e3e145dbba17219be86d157ba6ebf3b5c25c8f7ba2d5723986fb9b1. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration:1:cobol
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
Code Migration: Overall (Show measurement, test conditions, and source)
Source value
47.434
Score
47.434
Metric
accuracy
Unit
%
Benchmark version
1
Category
coding
Task
overall
Task label
Overall
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-29. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 7994.848. Standard error: 4.272. Cost per task in USD: 3.405133. Source ID: vals-ai. Evidence capture ID: vals-code-migration-2026-11-07-23c368426e3e. Source payload hash: 23c368426e3e145dbba17219be86d157ba6ebf3b5c25c8f7ba2d5723986fb9b1. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-code-migration:1:overall
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
CyberBench v1.1: Overall (Show measurement, test conditions, and source)
Source value
65.357
Score
65.357
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
cybersecurity
Task
overall
Task label
Overall
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
active
Retrieved at
2026-11-07
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-11-05. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 2202.787. Standard error: 5.55. Cost per task in USD: 0.457872. Source ID: vals-ai. Evidence capture ID: vals-cyber-2026-11-07-cc5fe09bf1f8. Source payload hash: cc5fe09bf1f8c926613f8893f823a902fb33f219303590d6a2794e43708d67a5. Parser version: vals-astro-v4. Imported at: 2026-11-07. Source cohort: vals-cyberbench:1.1:overall
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Source Vals AI
CyberBench v1.1: Overall (Show measurement, test conditions, and source)
Source value
65.357
Score
65.357
Metric
accuracy
Unit
%
Benchmark version
1.1
Category
cybersecurity
Task
overall
Task label
Overall
Direction
Higher is better
Source type
independent-evaluator
Evaluator
Vals AI
Status
superseded
Retrieved at
2026-10-16
Methodology
Source model ID: tencent/hy4-preview. Source model name: Hy4 Preview. Provider reported by source: Tencent. Evaluator: Vals AI. Dataset type: private. Leaderboard date: 2026-10-12. Temperature: 0.9. Top P: 1. Maximum output tokens: 64000. Latency in seconds: 2202.787. Standard error: 5.55. Cost per task in USD: 0.493575. Source ID: vals-ai. Evidence capture ID: vals-cyber-2026-10-16-943f593f929f. Source payload hash: 943f593f929f8971a6b746269a2851b5febc2aad4e2f1066895f715033cdc7f1. Parser version: vals-astro-v3. Imported at: 2026-10-16. Source cohort: vals-cyberbench:1.1:overall
Comparison methodology
Temperature: 0.9. Top P: 1. Maximum output tokens: 64000
Context
This result is retained as historical evidence and is not interchangeable with current results.
Source Vals AI

Page 1 of 34

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

The 100 most popular models of this category, 0 to 100, higher is more popular

20 of 101 models

GPT-6 Astra83
Claude Haiku 5.577
GPT-6 Sol73
GPT-573
DeepSeek V4.1 Flash 101769
Claude Opus 5.568
Qwen 3.8 27B68
Grok 4.767
Kimi K367
Qwen 3.8 Max 090266
Claude Sonnet 564
Qwen3.5 4B64
DeepSeek-V4-Flash64
Claude Fable 5.164
GPT-5.6 Terra64
MiniCPM-V 4.7 35B-A3B63
GPT-6.1 Sol63
Claude Opus 563
Qwen 3.8 Flash62
Hy4 preview39

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 7, 2026

Hy4 previewGemma 3n E4B IT 39AliceAI Foundation 80B A3B Base 39Command R7B 39Devstral Small 2 39

Method changes in the timeline, a jump at these points is no market event:

  1. 1November 1, 2026: Value only with search or usage data, 30-day attention

What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.

Similar models

More models of the same kind as Hy4 preview, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Parameters
770 billion
Source: Parameters
Active parameters
49 billion
Source: Active parameters

Input and output

Context window
1,024,000 tokens
Source: Context window
Maximum output
64,000 tokens
Source: Maximum output
Inputs
Text
Source: Inputs
Outputs
Text
Source: Outputs
Reasoning
Yes
Source: Reasoning
Tool calling
Yes
Source: Tool calling

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking
Unknown
no source
Watermarking technique
Not known
no source

Licence and access

Weights
Open weights
Source: Weights
Access
API
Source: Access
License
Apache 2.0
Source: License

Prices

No Tencent subscription includes this model. Through other providers from $1 a month (Command Code Go, with limited credits). 10 plans in Germany.

PlanMonthly list priceNote
Command Code Go$1with limited credit
Poe Basic$4.99with limited credit
Command Code GOAT$10with limited credit
Command Code Pro$20with limited credit
Poe Plus$19.99with limited credit
Poe Pro$49.99with limited credit
Command Code Max 10×$100with limited credit
Poe Advanced$99.99with limited credit
Command Code Max 20×$200with limited credit
Poe Max$249.99with limited credit

All plans in the AI subscription comparison

Frequently asked questions about Hy4 preview

How large is the context window of Hy4 preview?

1.02 M tokens. That is how much text Hy4 preview reads at once, about 1,540 A4 pages.

Page count as a rule of thumb from Artificial Analysis

Context windowTokens the model reads at once · higher is better
GPT-4.11.05M
GPT-4.1 mini1.05M
Hy4 previewRank 31 of 1031.02M
Claude Opus 5.51M
Gemini 4 Carbon1M
Claude Fable 5.11M
Claude Sonnet 5.51M

Can I run Hy4 preview locally?

Yes, the weights are open. With 770B parameters it needs a lot of memory, more a server than a laptop.

Stated by Tencent

Open weights
yes
License
Apache 2.0
Parameters
770B
active per token
49B

Where can I use Hy4 preview?

Not in a Tencent subscription. Including other providers that makes 10 plans in Germany.

Subscription comparison, prices in Germany incl. VAT

At Tencent

No Tencent subscription includes this model.

Through other providers

10 plans from 2 providers, from $1 a month (Command Code Go), with limited credits.

Sources and data date

Show all 42 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

EvaluatorTest seriesComposite indicesRecorded active measurements
Vals AI254262

Without vendor claims and historical or disputed results; those stay labelled in the single values.

TypeEvidence and data date
Input

Documented modalities. Retrieved October 27, 2026.

Source: models.dev catalog · models.dev catalog (retrieved October 27, 2026)
Output

Documented modalities. Retrieved October 27, 2026.

Source: models.dev catalog · models.dev catalog (retrieved October 27, 2026)
Research statusResearch date October 15, 2026. 8 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published

leaderboard-run-date-and-settings

Note

Tencent's exact-model card exposes independent leaderboard widgets but does not provide prompt/agent setup, inference effort, quantization, score-unit details for most rows, or task run dates.

Additional sourceHugging Face (retrieved October 15, 2026) · Tencent Hy4 preview model card · Editorial description reviewed October 2, 2026
Additional sourceVals AI (retrieved November 7, 2026)12345678910111213141516171819
Additional sourceVals AI (retrieved October 23, 2026; November 7, 2026)12
Additional sourceVals AI (retrieved October 16, 2026; November 7, 2026)1234
Additional sourceVals AI (retrieved October 16, 2026; October 31, 2026; November 7, 2026)
Additional sourceVals AI (retrieved October 16, 2026; October 30, 2026; October 31, 2026)
Additional sourceVals AI (retrieved October 28, 2026; November 7, 2026)
Additional sourceTencent, Hy4 preview model card, Model Introduction (retrieved October 17, 2026; October 19, 2026)
Additional sourceTencent (retrieved October 17, 2026; October 23, 2026) · Tencent Hy4 preview announcement · Editorial description reviewed October 2, 2026
Additional sourceDuelLab GameBench 2 model results (retrieved October 28, 2026; November 2, 2026; November 5, 2026)
Additional sourceMathBench.ai (retrieved October 28, 2026)
Additional sourceModel metadata (retrieved October 28, 2026)
Additional sourceTencent Hy4 preview configuration · Editorial description reviewed October 2, 2026
Additional sourceOpenRouter: Popularity index (derived), CC BY 4.0
Additional sourceVercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional sourceOllama: cumulative downloads for the popularity index
Additional sourceReplicate: cumulative runs for the popularity index
Additional sourceGoogle Trending Now: limited feed selection for Germany and the US
Additional sourceCreative Commons Attribution 4.0