ERNIE 4.5 VL 424B A47B
Unavailable in this round
Language modelActive
ERNIE 4.5 VL 424B A47B
Baidu
- Released
- June 30, 2025
- Data date
- September 8, 2026
- Price
- Input · Output / 1M tokens
- Input
- Text, Image
- Output
- Text
ERNIE 4.5 VL 424B A47B combines image understanding with a heterogeneous mixture-of-experts architecture. The open model accepts text and images, activates 47 of 424 billion parameters, and uses 54 layers with a 131,072-token context.
Baidu documents thinking and non-thinking modes for its vision-language post-training. Apache 2.0 applies to the published weights. The exact profile identifier ends in PT. In the model card, PT denotes the PyTorch format, not a pretraining or base model. That distinction keeps the suffix from being mistaken for a different model variant.
Specifications and access
| Specification | Value and source |
|---|---|
| Weights | Open weightsSource |
| Access | API, LocalSource |
| Parameters | 424BSource |
| Active parameters | 47BSource |
| Context window | 131,072 tokensSource |
| Release | 2025-06-30Source |
| License | Apache-2.0Source |
| Parameters | 424 BSource |
| Active parameters | 47 BSource |
| Architecture | multimodal heterogeneous sparse MoE, 54 layers, Q/KV heads 64/8, text experts 64/8, vision experts 64/8Source |
| Input | Text, ImageSource |
| Output | TextSource |
| Reasoning mode | thinking and non-thinking modes for vision-language post-trainingSource |
Page 1 of 4
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.
CCBench (thinking) (Show measurement, test conditions, and source)
- Source value
- 83.9
- Score
- 83.9
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
SimpleVQA (thinking) (Show measurement, test conditions, and source)
- Source value
- 59.8
- Score
- 59.8
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
MMBench-cn (thinking) (Show measurement, test conditions, and source)
- Source value
- 90.9
- Score
- 90.9
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
MMBench-en (thinking) (Show measurement, test conditions, and source)
- Source value
- 92
- Score
- 92
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
TableVQA (thinking) (Show measurement, test conditions, and source)
- Source value
- 86.7
- Score
- 86.7
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
ChartQA (thinking) (Show measurement, test conditions, and source)
- Source value
- 86.8
- Score
- 86.8
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
AI2D (thinking) (Show measurement, test conditions, and source)
- Source value
- 95.3
- Score
- 95.3
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
DocVQA (thinking) (Show measurement, test conditions, and source)
- Source value
- 93.1
- Score
- 93.1
- Metric
- percent
- Unit
- %
- Category
- vision
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
VisualPuzzle (thinking) (Show measurement, test conditions, and source)
- Source value
- 46.3
- Score
- 46.3
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
ZeroBench (sub) (thinking) (Show measurement, test conditions, and source)
- Source value
- 22.5
- Score
- 22.5
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
Visulogic (thinking) (Show measurement, test conditions, and source)
- Source value
- 27.3
- Score
- 27.3
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
MMMU (thinking) (Show measurement, test conditions, and source)
- Source value
- 70
- Score
- 70
- Metric
- percent
- Unit
- %
- Category
- reasoning
- Direction
- Higher is better
- Source type
- vendor-reported
- Evaluator
- Baidu ERNIE 4.5 official VLM thinking benchmark
- Status
- active
- Retrieved at
- 2026-09-08
- Methodology
- Vendor-reported result. Official VLM table, thinking mode; ERNIE-4.5-VL-424B-A47B column; exact shots and decoding parameters are not stated.
Page 1 of 4
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.
ERNIE 4.5 VL 424B A47B
Reasoning (requested): high
Current attempt
- Run date
- Oct 2, 2026
- API provider
- openrouter
- Requested API model
- baidu/ernie-4.5-vl-424b-a47b
- API model reported in response
- baidu/ernie-4.5-vl-424b-a47b
- Reasoning setting
- high
- Pinned endpoint
- Novita | baidu/ernie-4.5-vl-424b-a47b
- Temperature
- Provider default
- Token limit including reasoning
- 8,192 tokens
- Test protocol
- community-visual-tests-v1
- Technical checks
- Passed, not a quality rating
- Response time
- 40.1 s
Visually reviewed. This individual test does not establish overall model quality.
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
23.7 / 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. 6 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 ERNIE-4.5-VL card does not publish a knowledge cutoff. |
| Unresolved | AA benchmark rows Original wording No exact Artificial Analysis model page located in the checked public sitemap. |
| Additional source | Baidu ERNIE 4.5 official VLM thinking benchmark (retrieved September 8, 2026) |
| Additional source | Baidu ERNIE 4.5 official VLM non-thinking benchmark (retrieved September 8, 2026) |
| Additional source | Model metadata (retrieved September 8, 2026) · Baidu ERNIE 4.5 official announcement · 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 |