Tev1-4B-experimental
Tev1-4B-experimental by Together is a decision model: it answers typed questions with a probability for each allowed answer instead of generating text. It scores 34.1 on the Jev Decision Index (rank 98). Hosted, it costs about $0.0084 per 1,000 decisions.
Pricing
Cost per 1,000 decisions assumes 200 input tokens per decision.
Specifications
- Provider
- Together
- 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
- No
- Runs on CPU
- No
- License
- —
- Hugging Face
- togethercomputer/Tev1-4B-experimental
Benchmarks
- Jev Decision IndexECE 0.104
- 34.1
- DecisionBenchECE 0.111
- 0.582
- jev-bench (Jevify)
- 0.703
Jev Decision Index: public 29.5, ECE 0.104, Brier 0.476. Lower ECE means its probabilities can be taken at face value.
Speed & latency
Compare Tev1-4B-experimental with…
FAQ
How much does Tev1-4B-experimental cost?
Hosted, Tev1-4B-experimental costs $0.04 per 1M input tokens, about $0.0084 per 1,000 decisions at 200 input tokens each.
What questions can Tev1-4B-experimental answer?
Tev1-4B-experimental answers choice (one of n options), score (ordered rubric), yes / no questions, returning a probability for each allowed answer.
Is Tev1-4B-experimental well calibrated?
On the Jev Decision Index its expected calibration error (ECE) is 0.104; below about 0.05, a stated 80% confidence is right roughly 80% of the time.
Is Tev1-4B-experimental open weight, and can it run on a CPU?
Yes, Tev1-4B-experimental 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 (OpenRouter, Jev Decision Index, DecisionBench, Enterprise Decision Arena). Pricing is per 1M tokens (USD); confirm with the provider before production use.