Kakao Corp.Decision modelActive

Kanana-2-v-decide-3b-2611

Summary

Kakao released Kanana-2-v-decide-3b-2611 on November 6, 2026. It answers supplied choice, yes-or-no, and scoring questions with probabilities instead of free-form text. Inputs can be text, JSON, or chat messages, optionally with up to ten images; the model card lists up to 20,480 input tokens. Video is not supported.

The checkpoint has 3.67 billion parameters, including 0.49 billion in the vision encoder and 3.18 billion in the language model. The weights are available for local use under the Kanana Open License Agreement. That license requires a separate commercial license for third-party access through an API, cloud platform, or other remote service.

All benchmark results

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

Show all 3 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.

Kanana Decision Bench (Text) (Show measurement, test conditions, and source)

Source value: 72.5
Score: 72.5
Metric: points
Unit: points
Category: decision
Direction: Higher is better
Sample: 15909
Source type: vendor-reported
Evaluator: Kakao
Status: active
Retrieved at: 2026-11-08
Methodology: Vendor-reported result. Kakao's internal Korean text decision set.

Source Kakao

Decision Index 0.4.1 (Public) (Show measurement, test conditions, and source)

Source value: 61.34
Score: 61.34
Metric: index score
Unit: points
Benchmark version: 0.4.1
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Kakao
Status: active
Retrieved at: 2026-11-08
Methodology: Vendor-reported result. Public text benchmark as evaluated by Kakao.

Source Kakao

Decision Index Vision 0.4.1 (Public) (Show measurement, test conditions, and source)

Source value: 61.12
Score: 61.12
Metric: index score
Unit: points
Benchmark version: 0.4.1
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Kakao
Status: active
Retrieved at: 2026-11-08
Methodology: Vendor-reported result. Public vision benchmark as evaluated by Kakao.

Source Kakao

Similar models

More models of the same kind as Kanana-2-v-decide-3b-2611, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Model type: A decision model based on Kanana-2-v-3b, pairing the Kanana-2-3b language model with the Kanana vision encoder.; Source: Model type

Input and output

Inputs: Text, JSON, or chat messages, optionally with images.; Source: Inputs

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: Kakao published the weights in its official Hugging Face repository; the model card shows local use with Transformers.; Source: Access
License: The Kanana Open License Agreement applies to local use of the weights. A separate commercial license is required to offer third-party access through an API, cloud platform, or other remote service.; Source: License
Shutdown: None announced; no source

API

Model ID: kakaocorp/kanana-2-v-decide-3b-2611; Source: Model ID

Further details

Result: Returns a probability distribution for each supplied choice, yes-or-no, or scoring question. It does not generate free-form text.; Source: Result
Input limits: Up to 10 images and 20,480 input tokens.; Source: Input limits
Video input: Not supported.; Source: Video input
Parameters and embedding: 3.67 billion parameters: 0.49 billion in the vision encoder and 3.18 billion in the language model. Output embeddings have 2,560 dimensions.; Source: Parameters and embedding

Sources and data date

Benchmark sources at a glance

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

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

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

Type	Evidence and data date
Research status	Research date November 8, 2026. 1 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Additional source	Kakao Corp. Kanana-2-v-decide-3b-2611 model card (retrieved November 8, 2026) · Kakao Corp. Kanana-2-v-decide-3b-2611 model card · Editorial description reviewed November 8, 2026
Additional source	Kanana Open License Agreement (retrieved November 8, 2026) · Kanana Open License Agreement · Editorial description reviewed November 8, 2026
Additional source	Kakao (retrieved November 8, 2026)