Releases·Model Release
Aurora1.0-150M Releases: 7B-Token Model Matching GPT-2 Small
A developer released Aurora1.0-150M, a 150M-parameter model performing similarly to GPT-2-Small, on Hugging Face. It was trained on 7B tokens using an RTX Pro 6000, with an example inference script included in the repo. The release offers practitioners a compact baseline model and open benchmarks.
175 engagement1 sourceSun, Sep 13, 2026, 06:46 PM
AI Overview
- Aurora1.0-150M, trained on 7B tokens on an RTX Pro 6000, scores 62.24% PIQA and 32.20% HellaSwag, performance similar to GPT-2 Small
- Additional benchmarks: 44.91% Arc-Easy, 25.00% Arc-Challenge, 33.90% Arithmark 3.0, 36.55% CapitalBench; inference script available in the Hugging Face repo
Related
Releases·Model Release
Aurora-80K releases: an 80K-parameter tiny language model
Aurora-80K, a language model with exactly 80,000 parameters, has been released on Hugging Face. It uses a factorized 4,096-token vocabulary to achieve compact representation. Benchmarks show Wikitext-2 BPB of 3.2902, BLiMP 52.31%, and Arc-Easy 26.05%.
Aurora-80K, a language model with exactly 80,000 parameters, has been released on Hugging Face. It uses a factorized 4,096-token vocabulary to achieve compact representation. Benchmarks show Wikitext-2 BPB of 3.2902, BLiMP 52.31%, and Arc-Easy 26.05%.
246 engagement1 sourceredditThu, Aug 20, 2026, 11:45 AM
Releases·Model Release
Developer releases BetterGPT-150M, a 152M-parameter model trained on 15B tokens
A developer released BetterGPT-150M, a compact causal language model with ~152 million parameters trained on 15B tokens. Benchmarks show it outperforms GPT-2 Small while matching models trained on larger token budgets. A live Hugging Face Space demo is available.
A developer released BetterGPT-150M, a compact causal language model with ~152 million parameters trained on 15B tokens. Benchmarks show it outperforms GPT-2 Small while matching models trained on larger token budgets. A live Hugging Face Space demo is available.
1 engagement1 sourceredditWed, Jul 29, 2026, 06:47 AM
Releases·Model Release
Developer pre-trains Shibai-700M-Base on 18B tokens for Python and Wikitext
A developer released Shibai-700M-Base, a 700M parameter model pre-trained on 18B tokens of Python and Wikitext. The model is not fine-tuned for chat and is intended for continued pre-training on doc string Python. The author acknowledges it likely underperforms Qwen 3 0.6B and Qwen 3.5 0.8B but claims it is an order of magnitude better than GPT-2.
A developer released Shibai-700M-Base, a 700M parameter model pre-trained on 18B tokens of Python and Wikitext. The model is not fine-tuned for chat and is intended for continued pre-training on doc string Python. The author acknowledges it likely underperforms Qwen 3 0.6B and Qwen 3.5 0.8B but claims it is an order of magnitude better than GPT-2.
61 engagement1 sourceredditWed, Jul 29, 2026, 07:29 PM
Releases·Model Release
Community skeptical of endless-frontier/BigBang-v1 Qwen 3.5 finetune benchmarks
A Reddit user shares endless-frontier/BigBang-v1, a Qwen 3.5 finetune converted to GGUF by Bartowski, but expresses doubts about its headline benchmark claims. The user notes the performance claim is vague and questions the methodology behind the averaged benchmarks.
A Reddit user shares endless-frontier/BigBang-v1, a Qwen 3.5 finetune converted to GGUF by Bartowski, but expresses doubts about its headline benchmark claims. The user notes the performance claim is vague and questions the methodology behind the averaged benchmarks.
37 engagement1 sourceredditSun, Aug 9, 2026, 09:22 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