TencentLanguage modelActive

Hy4

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

Summary

Hy4 is a new language model in Tencent's Hunyuan line, released on November 4, 2026. It understands images and can operate computers through both the command line and graphical interfaces. Tencent names complex chart analysis, software development, and multi-step agent tasks as use cases. Hy4 is a line of its own and not the same as Hy4 preview.

Through Tencent Cloud TokenHub the model offers a 1M-token context window, with at most 960k tokens of input and 64k of output. Prices are 6 CNY per million input tokens, 18 CNY per million output tokens, and 0.3 CNY for cache hits; TokenHub halves them through December 31, 2026. Tencent says weights and a system card will follow soon, and the announcement names no license.

All benchmark results

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

Show all 137 published benchmark results

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.

EMB v1.1: Comps (Show measurement, test conditions, and source)

Source value: 77.04
Score: 77.04
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: comps
Task label: Comps
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 984.992. Standard error: 3.837. Cost per task in USD: 0.781644. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:comps
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Dataroom (Show measurement, test conditions, and source)

Source value: 86.922
Score: 86.922
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: dataroom
Task label: Dataroom
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 891.787. Standard error: 3.517. Cost per task in USD: 0.623767. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:dataroom
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: DCF (Show measurement, test conditions, and source)

Source value: 56.642
Score: 56.642
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: dcf
Task label: DCF
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1039.185. Standard error: 8.425. Cost per task in USD: 0.619511. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:dcf
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: LBO (Show measurement, test conditions, and source)

Source value: 56.829
Score: 56.829
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: lbo
Task label: LBO
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1490.469. Standard error: 5.691. Cost per task in USD: 0.844803. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:lbo
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: M&A (Show measurement, test conditions, and source)

Source value: 60.114
Score: 60.114
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: ma
Task label: M&A
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 950.899. Standard error: 9.2. Cost per task in USD: 0.446947. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:ma
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Misc (Show measurement, test conditions, and source)

Source value: 57.679
Score: 57.679
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: misc
Task label: Misc
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1062.635. Standard error: 8.383. Cost per task in USD: 0.528321. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:misc
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Operating (Show measurement, test conditions, and source)

Source value: 68.024
Score: 68.024
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: operating
Task label: Operating
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 956.44. Standard error: 5.473. Cost per task in USD: 0.461407. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:operating
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Overall (Show measurement, test conditions, and source)

Source value: 66.178
Score: 66.178
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: overall
Task label: Overall
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1058.904. Standard error: 2.641. Cost per task in USD: 0.621653. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:overall
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Scratch (Show measurement, test conditions, and source)

Source value: 58.611
Score: 58.611
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: scratch
Task label: Scratch
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1078.479. Standard error: 4.136. Cost per task in USD: 0.51914. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:scratch
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

EMB v1.1: Template (Show measurement, test conditions, and source)

Source value: 74.211
Score: 74.211
Metric: accuracy
Unit: %
Benchmark version: 1.1
Category: finance
Task: template
Task label: Template
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1039.33. Standard error: 2.838. Cost per task in USD: 0.724167. Source ID: vals-ai. Evidence capture ID: vals-emb-v1-1-2026-11-06-fe7c4f471990. Source payload hash: fe7c4f47199098b80ba5ba70baf56593508c341ac801deeacf5d1325ab4eaafb. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-emb-v1-1:1.1:template
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

Finance Agent v2.1: Adjustments (Show measurement, test conditions, and source)

Source value: 64.632
Score: 64.632
Metric: accuracy
Unit: %
Benchmark version: 2.1
Category: finance
Task: adjustments
Task label: Adjustments
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 1969.327. Standard error: 6.658. Cost per task in USD: 1.137324. Source ID: vals-ai. Evidence capture ID: vals-fabv2-1-2026-11-06-0c62df1c4671. Source payload hash: 0c62df1c4671d91efa36e7597bae0e7e575080ba39b2d56b6b1780bf9519e53a. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-finance-agent-v2-1:2.1:adjustments
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

Finance Agent v2.1: All-Pass (Show measurement, test conditions, and source)

Source value: 52.579
Score: 52.579
Metric: accuracy
Unit: %
Benchmark version: 2.1
Category: finance
Task: all_pass
Task label: All-Pass
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Vals AI
Status: active
Retrieved at: 2026-11-06
Methodology: Source model ID: tencent/hy4. Source model name: Hy4. 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: 65536. Latency in seconds: 2238.942. Standard error: 2.208. Cost per task in USD: 1.277206. Source ID: vals-ai. Evidence capture ID: vals-fabv2-1-2026-11-06-0c62df1c4671. Source payload hash: 0c62df1c4671d91efa36e7597bae0e7e575080ba39b2d56b6b1780bf9519e53a. Parser version: vals-astro-v4. Imported at: 2026-11-06. Source cohort: vals-finance-agent-v2-1:2.1:all_pass
Comparison methodology: Temperature: 0.9. Top P: 1. Maximum output tokens: 65536

Source Vals AI

Page 1 of 12

All specifications

Input and output

Capabilities: Image understanding and computer use. Tencent cites complex chart analysis and computer interaction through command-line and graphical interfaces.; Source: Capabilities

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

Access: Available in Tencent Cloud TokenHub since November 4, 2026. Tencent also lists WorkBuddy, CodeBuddy, Yuanbao, the WeCom Enterprise Assistant, and ima.; Source: Access

API

API model ID: hy4; Source: API model ID

Further details

Model line: New Hunyuan language model, distinct from Hy4 preview. Tencent reports expanded pretraining and post-training, plus image understanding.; Source: Model line
Documented limits: Tencent Cloud TokenHub lists a 1M-token context window, up to 960k input tokens, and up to 64k output tokens.; Source: Documented limits
Weights and license: Tencent says model weights and a system card will be released soon. The announcement does not name a license.; Source: Weights and license

Prices

Input price: 6 CNY per 1M tokensSource
Output price: 18 CNY per 1M tokensSource
Cache-hit price: 0.3 CNY per 1M tokensSource
TokenHub promotional price: Tencent Cloud TokenHub charges half the list price through December 31, 2026.Source

Sources and data date

Show all 13 sources

Benchmark sources at a glance

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

Evaluator	Test series	Composite indices	Recorded active measurements
Vals AI	7	3	137

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

Type	Evidence and data date
Additional source	Tencent Hy4 announcement (retrieved November 5, 2026) · Tencent Hy4 announcement · Editorial description reviewed November 5, 2026
Additional source	Tencent Cloud TokenHub model list (retrieved November 5, 2026) · Tencent Cloud TokenHub model list · Editorial description reviewed November 5, 2026
Additional source	Tencent Cloud TokenHub product updates (retrieved November 5, 2026) · Tencent Cloud TokenHub product updates · Editorial description reviewed November 5, 2026
Additional source	Vals AI (retrieved November 6, 2026)1 2 3 4 5 6 7 8 9 10