lev

InterfazeDecision modelOpen weights

lev by Interfaze is a decision model: it answers typed questions with a probability for each allowed answer instead of generating text. It scores 36.5 on the Jev Decision Index (rank 94).

Pricing

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

Specifications

Provider: Interfaze
Parameters: 5B
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: Qwen/Qwen3.5-4B-Base
Jev-compatible request: Yes
Runs on CPU: No
License: apache-2.0
Hugging Face: interfaze-ai/lev

Benchmarks

Jev Decision IndexECE 0.064: 36.5
Benchmark Heaven JevBench (Capability): 61.7
PT-BR typed decisions: 61.8

Jev Decision Index: public 39.2, ECE 0.064, Brier 0.490. Lower ECE means its probabilities can be taken at face value.

Speed & latency

Median latency per call (Decision Index)

71 ms

Compare lev with…

lev vs amos-1-large—/1M in lev vs amos-1-medium—/1M in lev vs Celeris-1 Decision$0.04/1M in lev vs ScaleDown Classify-1$0.04/1M in lev vs Cloudflare clef—/1M in lev vs Cloudflare clef-flash—/1M in

FAQ

How much does lev cost?

lev has no hosted price listed; its weights are open, so the cost is the compute you run it on.

What questions can lev answer?

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

Is lev well calibrated?

On the Jev Decision Index its expected calibration error (ECE) is 0.064; below about 0.05, a stated 80% confidence is right roughly 80% of the time.

Is lev open weight, and can it run on a CPU?

Yes, lev 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, Enterprise Decision Arena, Hugging Face). Pricing is per 1M tokens (USD); confirm with the provider before production use.