bongard-mini
bongard-mini by AgentBull is a decision model: it answers typed questions with a probability for each allowed answer instead of generating text.
Pricing
No hosted price is listed. The weights are open: the cost is your own compute.
Specifications
- Provider
- AgentBull
- Parameters
- 8B
- 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/t5gemma-2-4b-4b
- Jev-compatible request
- Yes
- Runs on CPU
- No
- License
- gemma
- Hugging Face
- AgentBull/bongard-mini
Benchmarks
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.
- Typed Decisions (agreement with teacher)Self-reportedTyped Decisions test accuracy vs soft gold (zero-shot; 400 cases / 2,000 decisions) · ECE 0.067
- 0.594
- DecisionBench 1.0 (23,900 decisions)Self-reported
- 78.0
- ImajevBench v2.0-lite (dev + calibration, 254 public-gold itSelf-reported
- 66.9
- typed-decisions (400 cases, 2,000 decisions)Self-reported
- 59.4
Compare bongard-mini with…
FAQ
How much does bongard-mini cost?
bongard-mini has no hosted price listed; its weights are open, so the cost is the compute you run it on.
What questions can bongard-mini answer?
bongard-mini answers choice (one of n options), score (ordered rubric), yes / no questions, returning a probability for each allowed answer.
Is bongard-mini well calibrated?
No calibration measurement for bongard-mini is recorded yet.
Is bongard-mini open weight, and can it run on a CPU?
Yes, bongard-mini 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 (Hugging Face). Pricing is per 1M tokens (USD); confirm with the provider before production use.