Decider-4B

Last updated:

Open5–35ms$0/M input

Quadrant scores

See the full quadrant

Scored with the public quadrant rubric (maturity × capability; bubble = adoption). Revise as evidence lands.

  • Maturity6.8/10

    Open weights under Apache-2.0, full HF documentation, TypeSafe API-compatible (POST /v1/systemone), reproducible training docs, per-type temperature maps — no production SLA keeps it below commercial tiers.

  • Capability7.3/10

    Official boards (17 Oct): Decision Index 0.3.2 Full 40.33 (#94); JevBench Intelligence 40.1, calibration 89.6, p50 0.025 s.

  • Adoption24/100

    Inherits momentum from decider-2b wave; v2.1 release went viral as first open model to beat Jev; still niche outside S1-focused builders.

Vendor claims

Board audit of 17 Oct 2026: see the “Board audit” section below; where it conflicts with older board statements on this page, it wins.

Decider-4B is the first open model to beat Jev on JevBench — a 4.2B dense model from Mapika with no reinforcement-learning stage, achieving rank #1 via superior speed (92.9) and cost efficiency (60.9), not raw intelligence (49.4 vs Jev’s higher reasoning scores).

Built on Qwen3.5-4B-Base (32 layers, 8 full attention + 24 gated delta-net linear attention), it reads shared state plus typed questions (choice, score, noul → catalog boolean) and returns calibrated distributions in one forward pass. Wire format targets TypeSafe POST /v1/systemone.

JevBench v1.4.2 Performance

Metric Score Notes
Overall 64.13 Rank #1 — first open to beat Jev
Intelligence 49.4 Below Jev on raw reasoning
Calibration 75.0 Good probability calibration
Speed 92.9 5.2ms with CUDA graphs
Cost ~60.9 ~$0.03/1k decisions

How it won: Decider-4B beats Jev on the composite JevBench score by being dramatically faster and cheaper to run, not by being smarter. On hard reasoning items, Jev still leads — but for production workloads where speed and cost matter, Decider-4B offers a compelling open alternative.

Specs

Attribute Value
Base model Qwen/Qwen3.5-4B-Base
Parameters 4.2B (dense)
Architecture 32 layers: 8 full attention + 24 gated delta-net linear attention
Weights 8.4 GB bf16
Current version v2.1 (2026-09-24)
License Apache-2.0
Stage 1 Cross-entropy on mixture v2 (742M tokens)
Stage 2 LoRA rank 64 + 29,325 harder decisions (merged)
RL stage None (unlike decider-2b)
Latency 5.2ms (CUDA graphs) / 32ms (eager) on B300
API POST /v1/systemone (TypeSafe-compatible)
Decision types choice, score, noul → boolean

Training (two stages, no RL)

Stage 1 (v1): One supervised pass over mixture v2 — ~95 public decision datasets, agent trajectories, Mind2Web, 26 additional public datasets, and 10 programmatic families with verifiable gold. 1.89M items, 742M tokens. AdamW on bf16 parameters directly (no FP32 master copy — this preserves more base-model knowledge).

Stage 2 (v2.1): LoRA rank 64 (alpha 128) on attention and MLP weights, 2 epochs over 29,325 rows of harder decisions. Key innovation: replay rows trained toward v1’s own answer distribution (KL divergence loss), not hard labels. This preserves sampled-play performance that v2 lost. Merged into bf16 weights after training.

No reinforcement-learning stage — unlike decider-2b, the 4B model skips RLCD entirely and still achieves top JevBench scores via efficient supervised training.

The Decider Family

Model Base Weights Use case
decider-0.8b Qwen3.5-0.8B 1.4 GB Smallest: routing, yes/no, short-state lookups
decider-2b Qwen3.5-2B 3.8 GB Default: routing, classification, browser agents
decider-4b (this) Qwen3.5-4B 8.4 GB Middle point: harder judgments, no RL stage
decider-35b-a3b Qwen3.5-35B-A3B (3B active) 65 GB When accuracy is worth 3–4× cost
decider-2b-vision Qwen3.5-2B VL 4.1 GB Decisions from images

All share one interface (decider.infer.Decider, POST /v1/systemone) and one readout mechanism.

Usage

from decider.infer import Decider
d = Decider("Mapika/decider-4b")

d.decide(
    "My card was charged twice for the same purchase.",
    [
        {"question": "Which department should handle this?",
         "options": ["billing", "technical support", "sales"]},
        {"question": "Does this need a refund action?",
         "options": ["no", "yes"]}
    ]
)
# [{'choice': 'billing', 'confidence': ..., 'probs': {...}},
#  {'choice': 'yes', 'confidence': ..., 'probs': {...}}]

Requirements: torch, transformers>=5, flash-linear-attention (Triton kernels for Qwen3.5 linear-attention layers). The per-type temperatures need decider-ai 1.4.0 or later.

Version History

Version Date Changes
v2.1 2026-09-24 Replay trained toward v1’s distribution; per-type temperature map
v2 2026-09-24 Stage 2 LoRA on harder decisions; better on hard judgments, worse on sampled play
v1 2026-09-22 First release: mixture v2 on Qwen3.5-4B-Base

Which version to use:

  • v2.1 (default): sampled play, games, browser agents, hard decisions
  • v2: if you need better-calibrated confidence on hard multi-step items
  • v1: form filling, BabyAI-GoTo navigation, greedy bag-draw play

Limitations

  1. No RL stage: Stated beliefs about action outcomes were not trained against exact laws. On live browser tasks v2.1 is 93.2% sampled; decider-2b v10 with RL is 93.2% but 91.7% on held-out tasks.

  2. Overconfident on hard items: Calibration error 0.147 on held-out generated families (v2: 0.046) and 0.184 on JevBench hard tier (v2: 0.104). Don’t read 0.8 confidence as 80% accuracy on hard questions.

  3. Known regressions vs v1: Issue #9 form case still wrong; BabyAI-GoTo 0.19 vs v1’s 0.54; greedy bag-draw 9.4 points under v1.

  4. Below the 35B: 2.4/2.6 points under decider-35b-a3b on regression set; 2.7 points on JevBench hard tier.

  5. English-centric: Multilingual rows (XNLI, PAWS-X, MASSIVE, Belebele, XCOPA) are a small share of training data.

  6. Schema cache disabled: v2.1’s LoRA was trained only in plain state-first layout. DECIDER_SCHEMA_CACHE=1 won’t activate for this model.

  7. Per-type temps need 1.4.0: Older decider-ai versions use global temperature 1.099 for all answers (same argmax, different probabilities).

Evaluation Summary (v2.1)

Benchmark Result
Regression set (in-task / held-out) 0.831 / 0.784
JevBench public (easy / standard / hard) 1.000 / 0.986 / 0.649
Live MiniWoB++ (sampled / greedy) 93.2% / 93.8%
Bespoke suite (macro / micro) 0.756 / 0.765
Zero-shot games (sampled win rate) 26.9%
CliffWalking −13 (teacher optimal)

Full evaluation tables with confidence intervals are in the HuggingFace model card.

Board audit (17 Oct 2026)

Checked against the official boards on 17 Oct 2026: Decision Index 0.3.2 (data of 16 Oct, 19:00 UTC), JevBench and ImageJevBench live. The boards are separate sources and are never averaged. Older board statements and rankings further down this page are superseded where they conflict with this section.

Board Row Result
Decision Index 0.3.2 (16 Oct 19:00 UTC) Decider 4B Full 40.33 (#94, tie group 23); public 41.19 (#93); same-skills 37.19; new-domains 40.4
JevBench open weights (live 17 Oct) decider-4b v2 Capability 64.9; Intelligence 40.1; calibration 89.6; p50 0.025 s; $0.015/1k

Score update (audit, 17 Oct 2026). Capability 8.2 → 7.3, first full working (the earlier score had no stored breakdown) = types 10 × 0.30 + calibration 6 × 0.25 + latency 7 × 0.25 + accuracy 5 × 0.20 = 7.25 → 7.3. Types from decisionTypes; calibration from the JevBench axis 89.6 (≥90 → 7, 85–89.9 → 6, 75–84.9 → 5, 60–74.9 → 4, below 60 → 3; no author ECE counted); latency from the board’s self-hosted p50 0.025 s without p99 (≤0.25 s → 7, ≤0.5 s → 6, ≤1 s → 5, above → 3); accuracy Decision Index Full 40.33 ÷ 60.11 × 8 = 5.4, half-up. Maturity and adoption unchanged.

Hugging Face on 17 Oct (Mapika/decider-4b): 29,432 downloads, 29 likes, licence tag apache-2.0, last modified 2026-10-12.

Not roudone’s Decider-4B (21 Oct 2026)

This page is Mapika’s Decider-4B. A different model with the same name, roudone/Decider-4B (“BankJev”, Apache-2.0, a NeoHorse-style pointer head, published 15 Oct), appears on JevBench open weights (Capability 56.8, Intelligence 27.1, v1.6.36). Those rows are not this model’s and are not used in its scores.