Hunyuan A13B Instruct
Output withheld for safety.
Language modelActive
Hunyuan A13B Instruct
Tencent
- Released
- June 27, 2025
- Data date
- September 8, 2026
- Price
- Input · Output / 1M tokens
- Input
- Text
- Output
- Text
Hunyuan A13B Instruct separates fast and slow thinking in a sparse 80-billion-parameter model with about 13 billion active per token. Fine-grained mixture-of-experts routing and grouped-query attention keep active compute limited. For agents, the model card documents a specialized vLLM parser for tool calls and internal reasoning steps.
The official model card shows a 32,768-token default configuration. Tencent publishes the weights and documents deployment through vLLM, SGLang, and TensorRT-LLM.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 80BSource |
| Active parameters | 13BSource |
| Context window | 262,144 tokensSource |
| License | Tencent Hunyuan A13B licenseSource |
| Release | 2025-06-27 (Tencent Cloud's hosted hunyuan-a13b API launch is dated 2025-06-25; this record tracks the exact open-weight checkpoint release instead.)Source |
| Parameters | 80 BSource |
| Active parameters | 13 BSource |
| Architecture | fine-grained sparse MoE with GQA; hybrid fast and slow thinking modesSource |
| Default context window | 32,768 tokensSource |
| Input | TextSource |
| Output | TextSource |
| Reasoning mode | hybrid slow-thinking and fast-thinking modesSource |
| Tool calling | YesSource |
Page 1 of 2
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.
AIME24 (Show measurement, test conditions, and source)
- Source value
- 87.3
- Score
- 87.3
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
AIME25 (Show measurement, test conditions, and source)
- Source value
- 76.8
- Score
- 76.8
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
MATH (technical-report table) (Show measurement, test conditions, and source)
- Source value
- 94.3
- Score
- 94.3
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; distinct report/evaluation setup from the first table; exact configuration not stated.
- Context
- Distinct table/setup from the 72.35 base-model table; do not merge silently.
GPQA-Diamond (Show measurement, test conditions, and source)
- Source value
- 71.2
- Score
- 71.2
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
OlympiadBench (Show measurement, test conditions, and source)
- Source value
- 82.7
- Score
- 82.7
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
LiveCodeBench (Show measurement, test conditions, and source)
- Source value
- 63.9
- Score
- 63.9
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
Fullstackbench (Show measurement, test conditions, and source)
- Source value
- 67.8
- Score
- 67.8
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
ArtifactsBench (Show measurement, test conditions, and source)
- Source value
- 43
- Score
- 43
- Metric
- percent
- Unit
- %
- Category
- coding
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
BBH (second table) (Show measurement, test conditions, and source)
- Source value
- 89.1
- Score
- 89.1
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; distinct evaluation setup; exact configuration not stated.
- Context
- Distinct table/setup from the 87.56 base-model table.
DROP (Show measurement, test conditions, and source)
- Source value
- 91.1
- Score
- 91.1
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
ZebraLogic (Show measurement, test conditions, and source)
- Source value
- 84.7
- Score
- 84.7
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
IF-Eval (Show measurement, test conditions, and source)
- Source value
- 84.7
- Score
- 84.7
- Metric
- percent
- Unit
- %
- Category
- instruction-following
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Tencent Hunyuan-A13B-Instruct model card
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Second benchmark table; exact shots and harness not stated.
Page 1 of 2
Our tests
SVG short test
One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.
Earlier tests
Pelican on a bicycle
A community classic for free-form SVG drawing.
Hunyuan A13B Instruct
Output withheld for safety
Reasoning (requested): high
Current attempt
- Run date
- Oct 5, 2026
- API provider
- openrouter
- Requested API model
- tencent/hunyuan-a13b-instruct
- API model reported in response
- tencent/hunyuan-a13b-instruct
- Reasoning setting
- high
- Pinned endpoint
- SiliconFlow | tencent/hunyuan-a13b-instruct
- Temperature
- Provider default
- Token limit including reasoning
- 8,192 tokens
- Test protocol
- community-visual-tests-v1
- Technical checks
- Failed
- Response time
- 24.2 s
Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.
Earlier attempts (2)
- Oct 2, 2026Output withheld for safetyReasoning setting: high
- Oct 1, 2026Output withheld for safetyReasoning setting: high
Prompt and test conditions
Generate an SVG of a pelican riding a bicycle
Token limit including reasoning: 8,192 tokens
Task origin (Simon Willison)Popularity
35.4 / 100
Sources and data date
| Type | Evidence and data date |
|---|---|
| Price | Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs. |
| Input | Documented modalities. Retrieved September 8, 2026. |
| Output | Documented modalities. Retrieved September 8, 2026. |
| Research status | Research date September 8, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Not found in the inspected sources | knowledge-cutoff Original wording The exact Hunyuan-A13B-Instruct card does not publish a training-data cutoff. |
| Unresolved | AA benchmark rows Original wording No exact Artificial Analysis model page located in the checked public sitemap. |
| Additional source | Model metadata (retrieved September 8, 2026) · Tencent Hunyuan A13B repository · Editorial description reviewed October 2, 2026 |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Creative Commons Attribution 4.0 |