MapikaDecision modelOpen source

Decider 35B A3B

PopularityRank 392 of 631

34out of 100-3 vs. last month

based on two data groups, mostly usage

Summary

Decider 35B A3B is the large MoE checkpoint in the Decider family. It contains 34.7 billion parameters, with roughly 3 billion active per token. Over a 32,000-token context, it answers typed text questions in one pass and returns probabilities for your options. Sparse activation still leaves it with the memory footprint of a large local model.

The quantized variant Decider 35B A3B NVFP4 keeps the same decision interface while reducing memory use.

All benchmark results

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

Show all 153 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.

Public regression in-task set (Show measurement, test conditions, and source)

Source value: 0.855
Score: 0.855
Metric: accuracy
Unit: ratio
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Mapika Decider results
Status: active
Retrieved at: 2026-10-15
Methodology: Vendor-reported result. BF16 decider-35b-a3b v1; public regression set; 67 in-task tasks.

Source Mapika Decider results

Public regression held-out set (Show measurement, test conditions, and source)

Source value: 0.81
Score: 0.81
Metric: accuracy
Unit: ratio
Category: decision
Direction: Higher is better
Source type: vendor-reported
Evaluator: Mapika Decider results
Status: active
Retrieved at: 2026-10-15
Methodology: Vendor-reported result. BF16 decider-35b-a3b v1; public regression set; 28 held-out tasks.

Source Mapika Decider results

Public regression in-task set ECE (Show measurement, test conditions, and source)

Source value: 0.026
Score: 0.026
Metric: expected calibration error (ECE)
Unit: ratio
Benchmark version: Mapika/decider MODEL_CARD_35B.md; repository revision d6fc522aed7fc392576a9b66a330409bc6c382bd; Decider-35B-A3B v1 BF16
Direction: Lower is better
Source type: vendor-reported
Evaluator: Mapika, Decider 35B-A3B model card
Status: active
Retrieved at: 2026-10-16
Methodology: Vendor-reported result. Decider-35B-A3B v1 BF16; one temperature fitted on the in-task set; ECE uses 15 bins; official card revision d6fc522aed7fc392576a9b66a330409bc6c382bd.
Context: The result belongs to the BF16 v1 checkpoint and the card's stated calibration recipe. It is not the NVFP4 derivative or a current serving API measurement.

Source Mapika, Decider 35B-A3B model card

Public regression in-task set negative log-likelihood (Show measurement, test conditions, and source)

Source value: 0.357
Score: 0.357
Metric: negative log-likelihood (nats)
Unit: No unit provided
Benchmark version: Mapika/decider MODEL_CARD_35B.md; repository revision d6fc522aed7fc392576a9b66a330409bc6c382bd; Decider-35B-A3B v1 BF16
Direction: Lower is better
Source type: vendor-reported
Evaluator: Mapika, Decider 35B-A3B model card
Status: active
Retrieved at: 2026-10-16
Methodology: Vendor-reported result. Decider-35B-A3B v1 BF16; one temperature fitted on the in-task set; ECE uses 15 bins; official card revision d6fc522aed7fc392576a9b66a330409bc6c382bd.
Context: The result belongs to the BF16 v1 checkpoint and the card's stated calibration recipe. It is not the NVFP4 derivative or a current serving API measurement.

Source Mapika, Decider 35B-A3B model card

Public regression held-out set ECE (Show measurement, test conditions, and source)

Source value: 0.069
Score: 0.069
Metric: expected calibration error (ECE)
Unit: ratio
Benchmark version: Mapika/decider MODEL_CARD_35B.md; repository revision d6fc522aed7fc392576a9b66a330409bc6c382bd; Decider-35B-A3B v1 BF16
Direction: Lower is better
Source type: vendor-reported
Evaluator: Mapika, Decider 35B-A3B model card
Status: active
Retrieved at: 2026-10-16
Methodology: Vendor-reported result. Decider-35B-A3B v1 BF16; one temperature fitted on the in-task set; ECE uses 15 bins; official card revision d6fc522aed7fc392576a9b66a330409bc6c382bd.
Context: The result belongs to the BF16 v1 checkpoint and the card's stated calibration recipe. It is not the NVFP4 derivative or a current serving API measurement.

Source Mapika, Decider 35B-A3B model card

Public regression held-out set negative log-likelihood (Show measurement, test conditions, and source)

Source value: 0.497
Score: 0.497
Metric: negative log-likelihood (nats)
Unit: No unit provided
Benchmark version: Mapika/decider MODEL_CARD_35B.md; repository revision d6fc522aed7fc392576a9b66a330409bc6c382bd; Decider-35B-A3B v1 BF16
Direction: Lower is better
Source type: vendor-reported
Evaluator: Mapika, Decider 35B-A3B model card
Status: active
Retrieved at: 2026-10-16
Methodology: Vendor-reported result. Decider-35B-A3B v1 BF16; one temperature fitted on the in-task set; ECE uses 15 bins; official card revision d6fc522aed7fc392576a9b66a330409bc6c382bd.
Context: The result belongs to the BF16 v1 checkpoint and the card's stated calibration recipe. It is not the NVFP4 derivative or a current serving API measurement.

Source Mapika, Decider 35B-A3B model card

JevBench hard (Show measurement, test conditions, and source)

Source value: 0.676
Score: 0.676
Metric: accuracy
Unit: ratio
Benchmark version: Mapika/decider MODEL_CARD_35B.md; repository revision d6fc522aed7fc392576a9b66a330409bc6c382bd; Decider-35B-A3B v1 BF16
Direction: Higher is better
Sample: 111
Source type: vendor-reported
Evaluator: Mapika, Decider 35B-A3B model card
Status: active
Retrieved at: 2026-10-16
Methodology: Vendor-reported result. Decider-35B-A3B v1 BF16; one temperature fitted on the in-task set; ECE uses 15 bins; official card revision d6fc522aed7fc392576a9b66a330409bc6c382bd.
Context: The result belongs to the BF16 v1 checkpoint and the card's stated calibration recipe. It is not the NVFP4 derivative or a current serving API measurement.

Source Mapika, Decider 35B-A3B model card

DecisionBench all-requested primary accuracy (Show measurement, test conditions, and source)

Source value: 0.5611297071129707
Score: 0.5611297071129707
Metric: all-requested accuracy, unsupported and error rows count as misses
Unit: ratio
Benchmark version: Hanno DecisionBench 1.0; result commit e2d6a133cee91b9d8c6fbd076b21f32f3d75c81f; dataset b7c8107e01ecb1aee7c7eaf5caee4a3ba9f59443; task spec a477586be2d19dc109c9f2479e3c455b76e6311758c05f42b5a0ca15f4c34b44; Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax
Category: decision
Direction: Higher is better
Sample: 23900
Source type: independent-evaluator
Evaluator: Hanno-Labs DecisionBench result repository
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; 34660610688 parameters; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax; Exact BF16 v1 checkpoint fine-tuned from Qwen/Qwen3.5-35B-A3B-Base; one public-mixture supervised epoch with routed experts frozen and Muon on block matrices; no RL stage. Submitted 2026-09-22T21:06:18.154638Z.
Context: This is one vendor-submitted run on the evaluator's benchmark, not a multi-seed estimate. No absolute-score confidence interval is reported.

Source Hanno-Labs DecisionBench result repository

DecisionBench supported-output accuracy (Show measurement, test conditions, and source)

Source value: 0.7295724077902296
Score: 0.7295724077902296
Metric: accuracy on successful supported outputs
Unit: ratio
Benchmark version: Hanno DecisionBench 1.0; result commit e2d6a133cee91b9d8c6fbd076b21f32f3d75c81f; dataset b7c8107e01ecb1aee7c7eaf5caee4a3ba9f59443; task spec a477586be2d19dc109c9f2479e3c455b76e6311758c05f42b5a0ca15f4c34b44; Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax
Category: decision
Direction: Higher is better
Sample: 18382
Source type: independent-evaluator
Evaluator: Hanno-Labs DecisionBench result repository
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; 34660610688 parameters; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax; Exact BF16 v1 checkpoint fine-tuned from Qwen/Qwen3.5-35B-A3B-Base; one public-mixture supervised epoch with routed experts frozen and Muon on block matrices; no RL stage. Submitted 2026-09-22T21:06:18.154638Z.
Context: This is one vendor-submitted run on the evaluator's benchmark, not a multi-seed estimate. No absolute-score confidence interval is reported.

Source Hanno-Labs DecisionBench result repository

DecisionBench output coverage (Show measurement, test conditions, and source)

Source value: 0.7691213389121339
Score: 0.7691213389121339
Metric: successful supported rows divided by requested rows
Unit: ratio
Benchmark version: Hanno DecisionBench 1.0; result commit e2d6a133cee91b9d8c6fbd076b21f32f3d75c81f; dataset b7c8107e01ecb1aee7c7eaf5caee4a3ba9f59443; task spec a477586be2d19dc109c9f2479e3c455b76e6311758c05f42b5a0ca15f4c34b44; Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax
Category: decision
Direction: Higher is better
Sample: 23900
Source type: independent-evaluator
Evaluator: Hanno-Labs DecisionBench result repository
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; 34660610688 parameters; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax; Exact BF16 v1 checkpoint fine-tuned from Qwen/Qwen3.5-35B-A3B-Base; one public-mixture supervised epoch with routed experts frozen and Muon on block matrices; no RL stage. Submitted 2026-09-22T21:06:18.154638Z.
Context: This is one vendor-submitted run on the evaluator's benchmark, not a multi-seed estimate. No absolute-score confidence interval is reported.

Source Hanno-Labs DecisionBench result repository

DecisionBench expected calibration error (Show measurement, test conditions, and source)

Source value: 0.044364205985813615
Score: 0.044364205985813615
Metric: 15-bin top-label expected calibration error
Unit: ratio
Benchmark version: Hanno DecisionBench 1.0; result commit e2d6a133cee91b9d8c6fbd076b21f32f3d75c81f; dataset b7c8107e01ecb1aee7c7eaf5caee4a3ba9f59443; task spec a477586be2d19dc109c9f2479e3c455b76e6311758c05f42b5a0ca15f4c34b44; Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax
Category: decision
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Hanno-Labs DecisionBench result repository
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; 34660610688 parameters; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax; Exact BF16 v1 checkpoint fine-tuned from Qwen/Qwen3.5-35B-A3B-Base; one public-mixture supervised epoch with routed experts frozen and Muon on block matrices; no RL stage. Submitted 2026-09-22T21:06:18.154638Z.
Context: The result JSON does not state a metric-specific N for ECE; none is inferred. One run and no confidence interval are reported.

Source Hanno-Labs DecisionBench result repository

DecisionBench mean negative log likelihood (Show measurement, test conditions, and source)

Source value: 0.8213308311369967
Score: 0.8213308311369967
Metric: mean negative log likelihood
Unit: No unit provided
Benchmark version: Hanno DecisionBench 1.0; result commit e2d6a133cee91b9d8c6fbd076b21f32f3d75c81f; dataset b7c8107e01ecb1aee7c7eaf5caee4a3ba9f59443; task spec a477586be2d19dc109c9f2479e3c455b76e6311758c05f42b5a0ca15f4c34b44; Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax
Category: decision
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Hanno-Labs DecisionBench result repository
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Mapika/decider-35b-a3b@d9783c58ad4fbff3a6030c8332d29738ec49d22a; 34660610688 parameters; adapter jev-compatible-server-v1; probability source option-slot-logit-softmax; Exact BF16 v1 checkpoint fine-tuned from Qwen/Qwen3.5-35B-A3B-Base; one public-mixture supervised epoch with routed experts frozen and Muon on block matrices; no RL stage. Submitted 2026-09-22T21:06:18.154638Z.
Context: The result JSON does not state the log base or metric-specific N; neither is inferred. One run and no confidence interval are reported.

Source Hanno-Labs DecisionBench result repository

Page 1 of 13

Popularity compared

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

Popularity index

0 to 100, higher is more popular

20 of 31 models

AutoTrust JEV-27B-VL70

Clef52

GLiNER2.5-multi-Decide45

Decider 4B43

Jev43

pplx-decider-v1.1-27b42

Intern-Decision-4B40

Kev 9B40

Laya Typed Decisions40

pplx-decider-v1-27b39

Decider 0.8B39

Julia 139

Intern-Decision-0.8B38

Decider 35B A3B34

GLiNER2.5-multi-Decide

Decider 4B

Jev

pplx-decider-v1.1-27b

Decider 35B A3B

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

Over time

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

Decider 35B A3BDecider 2B Vision 36Jeff 1 36GLiNER2.5-Decide-1B 34NanoJev 34

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

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 Decider 35B A3B, chosen by model line, developer, size, price and release date.

All specifications

Architecture

Model class: Typed decision model; Source: Model class
Parameters: 35 billion; Source: Parameters
Base model: Qwen3.5-35B-A3B-Base (3B aktiv); Source: Base model

Input and output

Input: Text and typed questions; Source: Input
Output: Structured probabilities, no text generation; Source: Output
Context window: 32,000 tokens; Source: Context window

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: Local; Source: Access
License: Apache-2.0; Source: License

API

Model ID: Mapika/decider-35b-a3b; Source: Model ID

Further details

Result: Choice, Noul and Score as probability distributions; Source: Result
Scope: Local readout without autoregressive text generation.; Source: Scope
Notes: The model is intended for structured decisions, not multi-step text answers.; Source: Notes
Choice options: 2-255 per question; Source: Choice options

Sources and data date

Show all 17 sources

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 October 15, 2026. 10 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	Exact model result in the public Decision Index 0.4.1 panel Note This index edition has no row that can be mapped to the profile repository/checkpoint exactly. This does not imply the model lacks results in other benchmarks.
Additional source	Decider repository (retrieved September 24, 2026) · Decider repository · Editorial description reviewed October 2, 2026
Additional source	Decider model card (retrieved September 24, 2026) · Decider model card · Editorial description reviewed October 2, 2026
Additional source	Decider 35B A3B checkpoint (retrieved September 24, 2026; October 17, 2026) · Decider 35B A3B checkpoint · Editorial description reviewed October 2, 2026
Additional source	Mapika Decider results (retrieved October 15, 2026; October 21, 2026)
Additional source	Mapika, Decider 35B-A3B model card (retrieved October 16, 2026)
Additional source	Hanno-Labs DecisionBench result repository (retrieved October 17, 2026)
Additional source	Mapika decider/docs/RESULTS.md (retrieved October 17, 2026)
Additional source	Mapika/decider RESULTS.md (external JevBench leaderboard recap) (retrieved October 19, 2026)
Additional source	Model metadata (retrieved October 17, 2026)1 2 3
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Ollama: cumulative downloads for the popularity index
Additional source	Replicate: cumulative runs for the popularity index
Additional source	Google Trending Now: limited feed selection for Germany and the US
Additional source	Creative Commons Attribution 4.0