Hopper 12B (trained)

HopitAIDecision modelOpen weights

Hopper 12B (trained) by HopitAI is a decision model: it answers typed questions with a probability for each allowed answer instead of generating text. It scores 47.3 on the Jev Decision Index (rank 57).

Pricing

No hosted price is listed. The weights are open: the cost is your own compute.

Specifications

Provider: HopitAI
Parameters: 12B
Released: Sep 2026
Open weights: Yes
Question types: Choice (one of N options), Score (ordered rubric), Yes / no
Mechanism: Decoder LM with a decision readout
Backbone: google/gemma-4-12B-it
Jev-compatible request: Yes
Runs on CPU: No
License: research-and-demo
Hugging Face: HopitAI/hopper

Benchmarks

Jev Decision IndexECE 0.076: 47.3
Jev Decision Indexedition 0.2 (earlier edition): 37.0
Benchmark Heaven JevBench (Capability): 68.2

Jev Decision Index: public 48.3, ECE 0.076, Brier 0.449. Lower ECE means its probabilities can be taken at face value.

Self-reported by the publisher

The model's own claims, often on its own sets or a self-run copy of a public suite. Not independently verified and never ranked alongside measured results.

Jev Decision Index public suite (self-run)Self-reportedDecision Index public balanced skill, edition 0.2.1 kit (release-v2.1), self-run: 47.5

Speed & latency

Median latency per call (Decision Index)

290 ms

Compare Hopper 12B (trained) with…

Hopper 12B (trained) vs amos-1-large—/1M in Hopper 12B (trained) vs amos-1-medium—/1M in Hopper 12B (trained) vs Celeris-1 Decision$0.04/1M in Hopper 12B (trained) vs ScaleDown Classify-1$0.04/1M in Hopper 12B (trained) vs Cloudflare clef—/1M in Hopper 12B (trained) vs Cloudflare clef-flash—/1M in

FAQ

How much does Hopper 12B (trained) cost?

Hopper 12B (trained) has no hosted price listed; its weights are open, so the cost is the compute you run it on.

What questions can Hopper 12B (trained) answer?

Hopper 12B (trained) answers choice (one of n options), score (ordered rubric), yes / no questions, returning a probability for each allowed answer.

Is Hopper 12B (trained) well calibrated?

On the Jev Decision Index its expected calibration error (ECE) is 0.076; below about 0.05, a stated 80% confidence is right roughly 80% of the time.

Is Hopper 12B (trained) open weight, and can it run on a CPU?

Yes, Hopper 12B (trained) has open weights. It needs a GPU to run at useful speed.

Data as of 2026-10-14. Source: Metatext decision-model catalogue, 2026-10-14 (Jev Decision Index, Benchmark Heaven JevBench, Hugging Face). Pricing is per 1M tokens (USD); confirm with the provider before production use.