Releases·Model Release
Next GenAi releases Rebekah AI 1.8B, an open-weight edge model under Apache 2.0
Next GenAi launched Rebekah AI 1.8B, an open-source, open-weight edge model under Apache 2.0, available on Hugging Face and Ollama. The company credits its cleaned, deduplicated, provenance-tracked Data Factory pipeline for competitive results against larger open models. It frames the release as the start of a broader Next GenAi edge model family.
1 engagement1 sourceTue, Sep 22, 2026, 08:48 AM
AI Overview
- Next GenAi released Rebekah AI 1.8B as an open-weight, open-source edge model on Hugging Face and Ollama under Apache 2.0
- Model scored 58.4% HumanEval, 62.8% GSM8K, 55.6% ARC Challenge, and 76.4% Systems & Tool Execution
- Next GenAi credits its Data Factory pipeline of cleaning, deduplicating, decontaminating, and provenance-tracking training data
- Developer claims the 1.8B model competes with open models 2-3x larger and positions it for edge and sovereign-AI deployment
Related
Social·Community
Alibaba open-weights Qwen3.8, a frontier model that beats Claude and GPT on some jobs
A Product Notes blog post argues Qwen3.8 is not simply the best model but something better, characterizing Alibaba's open-weight frontier model as beating Claude and GPT on some jobs while losing on others. The author describes how they are actually using the model in practice.
A Product Notes blog post argues Qwen3.8 is not simply the best model but something better, characterizing Alibaba's open-weight frontier model as beating Claude and GPT on some jobs while losing on others. The author describes how they are actually using the model in practice.
1.5k engagement2 sourcesrss, xThu, Sep 3, 2026, 07:28 AM
Releases·Tool Release
Spanda: Sub-microsecond LLM epistemic uncertainty in Rust
The Spanda team released an open-source Rust engine and gateway for LLM epistemic uncertainty quantification and hallucination gating, with a Python SDK (pip install spnda). It replaces the semantic entropy NLI cross-encoder bottleneck (90ms GPU overhead with DeBERTa) with exact-match normalized entropy, achieving equal AUROC on GSM8K in 760 nanoseconds on CPU. The approach makes real-time production hallucination gating feasible without GPUs.
The Spanda team released an open-source Rust engine and gateway for LLM epistemic uncertainty quantification and hallucination gating, with a Python SDK (pip install spnda). It replaces the semantic entropy NLI cross-encoder bottleneck (90ms GPU overhead with DeBERTa) with exact-match normalized entropy, achieving equal AUROC on GSM8K in 760 nanoseconds on CPU. The approach makes real-time production hallucination gating feasible without GPUs.
19 engagement1 sourcehackernewsFri, Sep 11, 2026, 06:48 PM
Releases·Model Release
Liquid AI releases LFM 2.5 QAD, a 2.6B quantized model
Liquid AI released LFM 2.5 QAD, a quantized version of its Liquid Foundation Model, with a 2.6B parameter GGUF file available on Hugging Face. The release was announced via a tweet from Liquid AI's official account.
Liquid AI released LFM 2.5 QAD, a quantized version of its Liquid Foundation Model, with a 2.6B parameter GGUF file available on Hugging Face. The release was announced via a tweet from Liquid AI's official account.
140 engagement1 sourceredditWed, Aug 19, 2026, 04:59 PM
Releases·Model Release
OpenBMB releases MiniCPM5-2B, tops Artificial Analysis index for ≤4B open models
OpenBMB released MiniCPM5-2B, scoring 15 on the Artificial Analysis Intelligence Index v4.2 — the highest of any open-weights model at or below 4B parameters. Weights are available on Hugging Face with code on GitHub, offering developers a capable small model option.
OpenBMB released MiniCPM5-2B, scoring 15 on the Artificial Analysis Intelligence Index v4.2 — the highest of any open-weights model at or below 4B parameters. Weights are available on Hugging Face with code on GitHub, offering developers a capable small model option.
402 engagement2 sourcesredditMon, Sep 7, 2026, 01:43 PM
Releases·Model Release
SupraLabs releases Supra2-Medium-Base, a 25M-parameter Qwen3 model
SupraLabs released Supra2-Medium-Base, a 25M-parameter model built on the Qwen3 architecture and trained from scratch on a new rig with RTX 5060 Ti 16GB and RTX 5060 8GB. The model reportedly competes with their previous 50M-parameter Supra-50M-Base in benchmarks, despite being half the size. It is a base model only; an instruction-tuned version may follow.
SupraLabs released Supra2-Medium-Base, a 25M-parameter model built on the Qwen3 architecture and trained from scratch on a new rig with RTX 5060 Ti 16GB and RTX 5060 8GB. The model reportedly competes with their previous 50M-parameter Supra-50M-Base in benchmarks, despite being half the size. It is a base model only; an instruction-tuned version may follow.
82 engagement1 sourceredditThu, Aug 20, 2026, 02:50 PM