Up to date
Mapika Apache 2.0 out 33 days ago
4.21B 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 29.8k · #424 of 619 tracked
Changed this week: 1 new quantized build on the Hub (11 in all).2026-10-24
Picked by rule from 2 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Decider is an open source .Net constraint programming solver for NP-complete problems, supporting integer variables, scheduling with cumulative constraints, and optimization.
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.
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: 11 quantized builds and 8 Spaces.
Top Spaces running it: multimodalart/jev-decision-index ♥ 802
Weights only, an estimate: leave room for the KV cache. Method
About 2.5 GB at INT4, 5.0 GB at FP8, 10 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-09-22), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.
18 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported Decision Index (DI) · Classify and make typed decisions | 51.02 | v3 | model author | 2026-10-23 |
| author-reported Decision Index (DI) · Classify and make typed decisions | 50.85 | decider.serve 1.10.0 | model author | 2026-10-23 |
| author-reported Fresh held-out yes/no set, scored the JevBench way | 44.3 | v3 | model author | 2026-10-23 |
| author-reported Image questions, mean of 11 sets | 0.89 | v3 | model author | 2026-10-23 |
| author-reported JevBench public items, hard tier | 0.64 | v3 | model author | 2026-10-23 |
| author-reported JevBench public items, standard tier | 0.99 | v3 | model author | 2026-10-23 |
| author-reported Visual7W / VSR (never trained on) | 0.873 / 0.833 | v3 | model author | 2026-10-23 |
The model returns a probability distribution over defined options in one forward pass without text generation.[1]
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.
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.
Listed in Decision models
| Pipeline tag | text-classification |
|---|---|
| Architectures | Qwen3_5ForCausalLM |
| Model type | qwen3_5_text |
| Dtype | bfloat16 |
| Params by dtype | BF16: 4,205,751,296 |
| Size band | M |
| MoE | no |
| Context length | 262,144 (config.json max_position_embeddings) |
| Languages | en |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
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.
Weights-only memory for decider-4b: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 10 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 5.0 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.5 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
| How it was made | full fine-tune (whole weights published) |
|---|---|
| Lineage bucket | F - fine-tune of an open model (public/synthetic data) |
| Lineage evidence | board file inspection (adapter/head/weight files) + HF base_model field |
| Base model | Qwen/Qwen3.5-4B-Base |
| DI 0.2.1 score | 40.70 (independent (DI 0.2.1 board)) |
| ECE (board) | 0.0837 |
| Median latency | 12.6 ms (DI board, 1x RTX PRO 6000, single process) |
| Hosted price (input) | self-hosted/M |
| Source | https://huggingface.co/Mapika/decider-4b |
Base model: Qwen/Qwen3.5-4B-Base
quantized: 11adapter: 1finetune: 6
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Mapika/decider-4b-GGUF | quantized | GGUF | — | — | apache-2.0 | Mapika | 3 | 9.9k |
| mradermacher/decider-4b-GGUF | quantized | GGUF | — | — | apache-2.0 | mradermacher | 2 | 1.9k |
| mindchain/decider-4b-v2-GGUF | quantized | GGUF | — | — | apache-2.0 | mindchain | 2 | 1k |
| midium-ai/decider-4b-dwq-4bit | quantized | -4bit | 4.21B | M | apache-2.0 | midium-ai | 2 | 184 |
| hayriyigit/decider-4b-tr | adapter | — | 4.21B | M | other | hayriyigit | 2 | 128 |
| llmtech/decider-4b-fp8 | quantized | -fp8 | 4.21B | M | apache-2.0 | llmtech | 2 | 71 |
| llmtech/decider-4b-nvfp4 | quantized | nvfp4 | 2.42B | M | apache-2.0 | llmtech | 1 | 72 |
| smolnikov/migom-4b | finetune | — | 4.21B | M | apache-2.0 | smolnikov | 1 | 63 |
| rovangju/decider-4b-W8A16 | quantized | — | 4.21B | M | apache-2.0 | rovangju | 0 | 58 |
| edihasaj/surebranch-4b | finetune | — | 4.21B | M | cc-by-sa-4.0 | edihasaj | 0 | 39 |
| seimanu/decider-4b-NPU2 | quantized | — | — | — | apache-2.0 | seimanu | 0 | 34 |
| Ar4ikov/decider-4b-AWQ-ASYM-W4A16 | quantized | AWQ | 4.21B | M | apache-2.0 | Ar4ikov | 0 | 23 |
| rovangju/decider-4b-W4A16 | quantized | -W4A16 | 4.21B | M | apache-2.0 | rovangju | 0 | 19 |
| PhillyLabs/philadelphia-decider-4b | finetune | — | 4.21B | M | apache-2.0 | PhillyLabs | 0 | 8 |
| ChiTako/decider-4b-automode | finetune | — | 4.21B | M | apache-2.0 | ChiTako | 0 | 7 |
| CommanderCuth/Heuristic-1 | finetune | — | — | — | mit | CommanderCuth | 0 | 0 |
| peonist-ai/halogen-npu-decider-4b | quantized | — | — | — | apache-2.0 | peonist-ai | 0 | 0 |
| hipfire-models/decider-4b | finetune | — | — | — | apache-2.0 | hipfire-models | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2026-10-20 | decider-4b v3: text and image decisions (vision/), Noul T 0.05; v2.1 under tag v2.1 | 585473d |
| 2026-10-12 | Card: MNLI accuracy on the full validation_matched split | 99d7fd5 |
| 2026-09-24 | decider-4b v2.1 | eb5fbdf |
| 2026-09-24 | Card: form-filling cases from GitHub issue #9 under regressions | 49564dd |
| 2026-09-24 | decider-4b v2 | 7ab294c |
| 2026-09-23 | Card: JevBench public items and speed measured again (GitHub b44b4c9) | c23ab4d |
| 2026-09-22 | decider-4b v1 | 5e10009 |
| 2026-09-22 | initial commit | e868ee0 |
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.
| Title | Source | Published | First seen | Signals |
|---|---|---|---|---|
| Strands Decider 2B: a small, open-source, decision model | strandsagents.com | 2026-10-01 | 2026-10-07 | hn |
| Decider - an easy to use, open source C# Constraint Solver | github.com | 2013-06-28 | 2026-10-07 | hn |
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
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
