Releases·Model Release
Mica v0.1 4B: open Jev-style decision model (yes/no, choice, score) that runs on an 8 GB GPU — trained for under $30 of GPU time
A developer released Mica v0.1 4B, an Apache-2.0 decision model for agent gates and routers that returns calibrated probabilities instead of generating text. Built on Qwen3.5-4B with a rank-16 LoRA and trained on 77,732 rows for under $30 of rented RTX 3090 time. It scores below the closed Jev 1.13 API on benchmarks but resists in-state prompt injection far better (69% vs 18%) and runs in about 47 ms per decision on an RTX 3090.
2 engagement1 sourceFri, Sep 25, 2026, 10:53 PM
AI Overview
- Sky7350's Mica v0.1 4B (Apache-2.0) is a Qwen3.5-4B LoRA fine-tune returning calibrated probabilities for yes/no, 2–255-option choices, or 2–10-level scores, compatible with the TypeSafe /v1/systemone format and trained on 77,732 rows (~34.7M tokens, half English, half Korean) for under $30 of RTX 3090 time.
- On a 7,328-decision held-out set, Mica scores 67.0 on English items vs Jev 1.13's 74.1, JevK5 4B's 61.0, and Kev 4B's 57.0, and resists in-state prompt injection (69% accuracy with a misleading note vs Jev's 18%).
- Mica ships as GGUF (Q4_K_M: 47 ms median latency on an RTX 3090 vs JevK5's 99 ms), reaches 69.5 on JevBench hard public items (64.9 via llama.cpp official runner), and shows weaknesses in knowledge tasks (MMLU-Pro 53.0 vs Jev's 82.3) and reversible-action judgment.
Related
Releases·Model Release
H2O.ai releases H2O-Lightning-4B, Apache-2.0 decision model that tops JevBench open leaderboard at 72.5 vs Jev 1.13's 71.5
H2O.ai released H2O-Lightning-4B, an open-weight Apache-2.0 model built for 'decision API' inference: it returns calibrated probabilities from a single forward pass with no generated tokens, making decisions fast and cheap. The 4B model (fine-tuned from Qwen3.5-4B) scores 72.5 on the JevBench public leaderboard, above Jev 1.13 at 71.5, making it the top open model. Announced by an H2O.ai employee with weights on Hugging Face; 12B and 31B versions are planned.
H2O.ai released H2O-Lightning-4B, an open-weight Apache-2.0 model built for 'decision API' inference: it returns calibrated probabilities from a single forward pass with no generated tokens, making decisions fast and cheap. The 4B model (fine-tuned from Qwen3.5-4B) scores 72.5 on the JevBench public leaderboard, above Jev 1.13 at 71.5, making it the top open model. Announced by an H2O.ai employee with weights on Hugging Face; 12B and 31B versions are planned.
77 engagement1 sourceredditFri, Oct 9, 2026, 08:26 PM
Products·launched
peekaboolean — 500M VLM answering typed image questions (choice/score/yes-no) with calibrated probabilities
peekaboolean is a small vision-language model that answers typed questions about a single image — multiple choice, rubric scoring, or yes/no — and returns calibrated probabilities instead of generated text. It solves the problem of slow, unparseable, poorly calibrated generative VLM outputs on local machines, for developers building apps that need fast image decisions. It runs about 400 ms per image on an M1 Pro without any text generation.
peekaboolean is a small vision-language model that answers typed questions about a single image — multiple choice, rubric scoring, or yes/no — and returns calibrated probabilities instead of generated text. It solves the problem of slow, unparseable, poorly calibrated generative VLM outputs on local machines, for developers building apps that need fast image decisions. It runs about 400 ms per image on an M1 Pro without any text generation.
1 engagement1 sourceredditSun, Sep 27, 2026, 08:46 AM
Releases·Model Release
EmpirioLabs releases Aplomb 1: open-weights 5.3B decision model, 1M context, text/image/video/audio in one request
EmpirioLabs released Aplomb 1, a 5.3B open-weights decision model built on Qwen3.5-4B that reads up to 1M tokens of text, JSON, images, video and audio in a single request. It returns probabilities for every tool and each enum/boolean argument, letting agents act on confident calls and defer uncertain ones to larger models. It ranks #1 on the Hugging Face Typed Decisions benchmark and #1 among 4B models on Decision Index 0.2.1.
EmpirioLabs released Aplomb 1, a 5.3B open-weights decision model built on Qwen3.5-4B that reads up to 1M tokens of text, JSON, images, video and audio in a single request. It returns probabilities for every tool and each enum/boolean argument, letting agents act on confident calls and defer uncertain ones to larger models. It ranks #1 on the Hugging Face Typed Decisions benchmark and #1 among 4B models on Decision Index 0.2.1.
Updated Fri, Oct 9, 2026, 03:58 PM — Aplomb 1 takes #1 on the Typed Decisions leaderboard with the best KL (0.123) and Brier (0.065) scores of any model list
02115 engagement3 sourcesredditTue, Oct 6, 2026, 05:56 PM
Products·launched
ARC-1: a 1.7B model for typed decisions (pick/score/yes-no) with probabilities
ARC-1 is a small decision model that takes context plus a question and returns a choice, score, or yes/no probability for tasks like ticket routing, intent detection, moderation, and agent tool calls. It is built by fine-tuning Qwen3-1.7B-Base with LoRA and runs in ~20 ms on a single consumer GPU. It targets developers building agentic or classification pipelines that need fast, cheap structured decisions.
ARC-1 is a small decision model that takes context plus a question and returns a choice, score, or yes/no probability for tasks like ticket routing, intent detection, moderation, and agent tool calls. It is built by fine-tuning Qwen3-1.7B-Base with LoRA and runs in ~20 ms on a single consumer GPU. It targets developers building agentic or classification pipelines that need fast, cheap structured decisions.
11 engagement1 sourceredditTue, Oct 6, 2026, 12:19 PM
Releases·Tool Release
DecisionTune 1.0: a 395M encoder that picks from your options offline, about 10 ms per short decision on MLX (Apache-2.0)
The model's developer released DecisionTune 1.0, an Apache-2.0 395M encoder (ModernBERT-large plus a 4 KB scoring head) that scores given options rather than generating text, aimed at routine agent decisions like tool routing and yes/no checks. It runs fully locally via PyTorch, MLX, or ONNX backends and ships with an MCP server for local assistants. The developer is seeking feedback on where it picks wrong.
The model's developer released DecisionTune 1.0, an Apache-2.0 395M encoder (ModernBERT-large plus a 4 KB scoring head) that scores given options rather than generating text, aimed at routine agent decisions like tool routing and yes/no checks. It runs fully locally via PyTorch, MLX, or ONNX backends and ships with an MCP server for local assistants. The developer is seeking feedback on where it picks wrong.
6 engagement1 sourceredditMon, Oct 5, 2026, 08:34 AM