Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › decider family

decider-2b

Hugging Face ↗ Follow changes

Mapika Apache 2.0 out 39 days ago

1.88B parameters, text in, labels out, 256K tokens of context.

Verified 2026-10-25 Rank #0 in Decision models · score 0 method v1 Sources 3 Downloads 30d 187.3k · #256 of 619 tracked

Can I run it?1.1GB at INT4Estimatefits a 16 GB cardmethod
Can I use it?Apache 2.0Permissive; patent grantlicence file2026-10-24
Do people use it?187.3kdownloads, 30 days#256 of 619 we track · top 41%+3 likes, downloads −44% this week110 likesHugging Face2026-10-24

Worth reading

Picked by rule from 2 tracked pieces: the lab's own post, the most shared, a field report, the latest.

Strands Decider 2B is version 19 of the architecture, having replaced a previous slot head that performed significantly worse.

From Strands Decider 2B: a small, open-source, decision model · strandsagents.com · 2026-10-01
Most shared · 2 sources

Strands Decider 2B: a small, open-source, decision model

Strands Labs released Strands Decider 2B, a 2 billion parameter decision model built on a Qwen3.5-2B torso with a rank-16 LoRA adapter.

  • The model architecture uses a pointer head that scores hidden states against the <answer> position to select options.
  • Strands Decider 2B is version 19 of the architecture, having replaced a previous slot head that performed significantly worse.
strandsagents.com 2026-10-01 236 HN pts · 67 comments 1.5k words

2 best of 2 tracked

The 1 pick above, plus these 1. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.

  1. CodeDecider - an easy to use, open source C# Constraint Solvergithub.com · 2013-06-28 · 1 HN pts

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

No API provider serves it, so its downloads are people running it themselves: 9 quantized builds and 14 Spaces.

  • As a service0no API provider listed
  • On own hardware9quantized builds
  • Built on107 fine-tunes · 3 adapters · 0 merges
  • In apps14Hugging Face Spaces

Top Spaces running it: multimodalart/jev-decision-index ♥ 802 · hugging-apps/decider-2b-demo ♥ 1

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 1.1 GB at INT4, 2.3 GB at FP8, 4.5 GB at BF16: weights only, so leave room for the KV cache.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 1.1 GB of weights (estimate; KV cache and activations extra)INT41.1 GBFP8: about 2.3 GB of weights (estimate; KV cache and activations extra)FP82.3 GBBF16: about 4.5 GB of weights (estimate; KV cache and activations extra)BF164.5 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 1.1 GB of weights (estimate; KV cache and activations extra)INT41.1 GBFP8: about 2.3 GB of weights (estimate; KV cache and activations extra)FP82.3 GBBF16: about 4.5 GB of weights (estimate; KV cache and activations extra)BF164.5 GB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

2 dated articles and 18 repo commits so far

Teal is the weights (2026-09-16), amber triangles are 2 dated articles (each links to it), grey ticks are repo commits.

JulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOct2026-10-20: 1 repo commit2026-10-12: 1 repo commit2026-09-24: 2 repo commits2026-09-23: 1 repo commit2026-09-22: 1 repo commit2026-09-20: 2 repo commits2026-09-19: 3 repo commits2026-09-18: 1 repo commit2026-09-17: 1 repo commit2026-09-16: 5 repo commitsStrands Decider 2B: a small, open-source, decision model · 2026-10-01Decider - an easy to use, open source C# Constraint Solver · 2013-06-28Weights 2026-09-16 · 2026-09-16Weights 2026-09-16now
JulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOctNovDecJanFebMarAprMayJunJulAugSepOct2026-10-20: 1 repo commit2026-10-12: 1 repo commit2026-09-24: 2 repo commits2026-09-23: 1 repo commit2026-09-22: 1 repo commit2026-09-20: 2 repo commits2026-09-19: 3 repo commits2026-09-18: 1 repo commit2026-09-17: 1 repo commit2026-09-16: 5 repo commitsStrands Decider 2B: a small, open-source, decision model · 2026-10-01Decider - an easy to use, open source C# Constraint Solver · 2013-06-28Weights 2026-09-16 · 2026-09-16Weights 2026-09-16now

What people built on it

19 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

19 derivatives on the Hub, led by 9 quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

9 quantized (Hugging Face model tree)7 fine-tunes (Hugging Face model tree)3 adapters (Hugging Face model tree)9 quantized7 fine-tunes3 adapters
9 quantized (Hugging Face model tree)7 fine-tunes (Hugging Face model tree)3 adapters (Hugging Face model tree)9 quantized7 fine-tunes3 adapters
GGUF Mapika/decider-2b-GGUF for llama.cpp, Ollama, LM Studio 4.9k downloads MLX edbordin-linktree/decider-2b-executorch-mlx for Apple silicon 22 downloads FP8 llmtech/decider-2b-fp8 for vLLM, SGLang 47 downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

14 rows

Benchmarks 14 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (14)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
BenchmarkValueSetting / notesReported byDate
author-reported Decision Index (DI) · Classify and make typed decisions38.17v12model author2026-10-23
author-reported Decision Index (DI) · Classify and make typed decisions38.18through decider.serve 1.10.0model author2026-10-23
author-reported Fresh held-out yes/no set, scored the JevBench way29.6v12model author2026-10-23
author-reported Image questions, mean of 11 sets0.85v12model author2026-10-23
author-reported Image questions, mean of 11 sets0.86trained kindsmodel author2026-10-23
author-reported Image questions, mean of 11 sets0.79never-trained (Visual7W, VSR)model author2026-10-23
author-reported Image questions, mean of 11 sets0.778 / 0.788v11 with the base vision tower and no image trainingmodel author2026-10-23
author-reported Image questions, mean of 11 sets0.77Stock Qwen3.5-2B with our readoutmodel author2026-10-23
author-reported JevBench public items1easymodel author2026-10-23
author-reported JevBench public items0.92standardmodel author2026-10-23
author-reported JevBench public items0.59hardmodel author2026-10-23
author-reported JevBench public items, hard tier0.59v12model author2026-10-23
author-reported JevBench public items, standard tier0.92v12model author2026-10-23
author-reported Visual7W / VSR (never trained on)0.863 / 0.717v12model author2026-10-23

What the card says

1 cited sentence

Summary generated, cited

The model returns a probability distribution over explicit option lists in one forward pass without text generation.[1]

  1. “returns a probability distribution over the options for e…”

Generated 2026-10-07 15:18 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

7 use cases

What it is good at 7 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

20 fields

Listed in Decision models

Facts

Pipeline tagtext-classification
ArchitecturesQwen3_5ForCausalLM
Model typeqwen3_5_text
Dtypebfloat16
Params by dtypeBF16: 1,881,825,088
Size bandS
MoEno
Context length262,144 (config.json max_position_embeddings)
Languagesen
LicenseApache 2.0 (apache-2.0)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3_5ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3_5ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for decider-2b: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)4.5 GB · Single GPU (<24 GB)
FP8 (1 byte/param)2.3 GB · Edge (<4 GB)
INT4 (0.5 bytes/param)1.1 GB · Edge (<4 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

Decision facts

catalog

Decision-model catalog curated

How it was madefull fine-tune (whole weights published)
Lineage bucketF - fine-tune of an open model (public/synthetic data)
Lineage evidenceboard file inspection (adapter/head/weight files) + HF base_model field
Base modelQwen/Qwen3.5-2B-Base
DI 0.2.1 score28.97 (independent (DI 0.2.1 board))
ECE (board)0.0771
Median latency8.1 ms (DI board, 1x RTX PRO 6000, single process)
Hosted price (input)self-hosted/M
Sourcehttps://huggingface.co/Mapika/decider-2b

All variants

19 variants

Lineage

Fine-tuned from Qwen/Qwen3.5-2B-Base Family: Mapika/decider 19 derivatives on the Hub

Base model: Qwen/Qwen3.5-2B-Base

quantized: 9finetune: 7adapter: 3

Top variants (19 shown, ranked by downloads)

Relation
Quant
19 rows
Variant Relation Quant Params Band License Author Likes Downloads
Mapika/decider-2b-GGUF quantized GGUF — — apache-2.0 Mapika 3 4.9k
cosetoenor/decider-2b-GGUF quantized GGUF — — apache-2.0 cosetoenor 8 2.3k
DreamBlooms/decider-2b-GGUF quantized GGUF — — apache-2.0 DreamBlooms 0 419
midium-ai/decider-2b-dwq-4bit quantized -4bit 1.88B S apache-2.0 midium-ai 2 206
nativ-community/decider-2b finetune — 1.88B S apache-2.0 nativ-community 0 140
Mapika/decider-2b-coherent finetune — — — apache-2.0 Mapika 1 82
tanakaisworking/Erabo-2B finetune — 1.88B S apache-2.0 tanakaisworking 1 55
llmtech/decider-2b-fp8 quantized -fp8 1.88B S apache-2.0 llmtech 1 47
johannhartmann/decider-2b-a11y-crawler-lora adapter — — — apache-2.0 johannhartmann 1 44
smolnikov/migom-2b finetune — 1.88B S apache-2.0 smolnikov 0 79
mlboydaisuke/decider-2b-coreai-ft finetune — 1.88B S apache-2.0 mlboydaisuke 0 52
daviestechlabs/decider-2b-homelab-router adapter — 1.88B S apache-2.0 daviestechlabs 0 44
seimanu/decider-2b-NPU2 quantized — — — apache-2.0 seimanu 0 33
johannhartmann/decider-2b-toolcall-guard-lora adapter — — — apache-2.0 johannhartmann 0 28
edbordin-linktree/decider-2b-executorch-mlx quantized mlx — — apache-2.0 edbordin-linktree 0 22
leafiy/mia-decider-2b finetune — — — apache-2.0 leafiy 0 0
goldenfox/decider-2b-onnx quantized — — — apache-2.0 goldenfox 0 0
ardana-ai/decider-2b-ONNX quantized — — — apache-2.0 ardana-ai 0 0
verdict-models/decider-2b finetune — — — apache-2.0 verdict-models 0 0

Repository changelog

18 commits
changed in the last 7 daysRepo commits: 2026-10-20all changes

Repository changelog 18 commits

DateCommitSHA
2026-10-20 decider-2b v12: text and image decisions (vision/), Noul T 0.05; v11 under tag v11 7614a6c
2026-10-12 Card: MNLI accuracy on the full validation_matched split ff99089
2026-09-24 decider-2b v11 533964d
2026-09-24 Card: field types for decider-ai 1.3.0 (confidence, x_p_max) d61c1c1
2026-09-23 Card: JevBench public items and speed measured again (GitHub b44b4c9) 9839cc9
2026-09-22 card: family table with decider-4b fa996ce
2026-09-20 card: the decider family table, contents, changelog b37f7e1
2026-09-20 card: point at decider-35b-a3b 0ef4a73
2026-09-19 Model card: v10 game and JevBench public-item results 4a0e867
2026-09-19 Upload folder using huggingface_hub 14d6555
2026-09-19 Update README.md 7789eb6
2026-09-18 Upload folder using huggingface_hub 1d96be0
2026-09-17 Upload folder using huggingface_hub f7d6447
2026-09-16 Upload folder using huggingface_hub 4195b2c
2026-09-16 Upload folder using huggingface_hub 760ab65
2026-09-16 Upload folder using huggingface_hub f5a4ed9
2026-09-16 Upload folder using huggingface_hub 4535722
2026-09-16 initial commit 51c6fbc

Articles and coverage

2 pieces

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Analysis2

Strands Decider 2B: a small, open-source, decision model
strandsagents.com2026-10-01HN front pageHN 236 pts · 67 comments1,464 words
Strands Labs released Strands Decider 2B, a 2 billion parameter decision model built on a Qwen3.5-2B torso with a rank-16 LoRA adapter. The model achieves a median latency of around 115ms on an Nvidia RTX3090 and ranks 3rd of 33 in the 2B class for accuracy and calibration. It removes the LM head to score options in parallel, making it suitable for model routing and guardrails where text generation is unnecessary.written, reviewed, cited
  • The model architecture uses a pointer head that scores hidden states against the <answer> position to select options.
  • Strands Decider 2B is version 19 of the architecture, having replaced a previous slot head that performed significantly worse.
  • The model can run locally on an M3 MacBook with a median latency for small tasks around 153ms.
Decider - an easy to use, open source C# Constraint Solver
github.com2013-06-28authorityabout the decider familyHN 1 pts · 0 comments782 words
Decider is an open source .Net constraint programming solver for NP-complete problems, supporting integer variables, scheduling with cumulative constraints, and optimization. Released under the MIT licence, it runs on .Net 10.0, 8.0, and Standard 2.0/2.1 while also supporting F#. Users can specify an upper time bound on how long you search using a CancellationToken.written, reviewed, cited

Similar models

3 models

Similar models

  • this-that-model-1.0 · 1.88B
    same size band (S), same task (text-classification), permissive license
  • this-that-model-1.2 · 1.88B
    same size band (S), same task (text-classification), permissive license
  • mojev · 854M
    same size band (S), same task (text-classification), permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:44 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:18 UTC · Last verified 2026-10-25

Follow decider-2b

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.