A wave of small and mid-size models, from 150M to 9.4B and MoE variants, shows competitive results with far lower hardware budgets. Data-efficiency methods and token merging reduce training and inference costs for on-premise deployments.
A Hugging Face page and online demo present a 230-million-parameter model fine-tuned for chess. The news is small, but it opens a practical discussion on specialized compact models for local deployments.
A 7B model in GGUF lands between 27B and 35B parameter models. For self-hosted deployment, K2 Horizon 7B shifts TCO calculations and makes VRAM efficiency the central metric, not parameter count. Yet deeper verification is needed.
A systematic evaluation across the Whisper family shows token merging boosts computational efficiency with almost no accuracy loss in multilingual transcription, even after DoRA fine-tuning. The method shortens token sequences during inference withou...
IFM's K2 Horizon 7B lands on the Artificial Analysis Intelligence Index between Qwen 3.6 27B and Qwen 3.6 35BA3b. In early local testing, it compiles the latest llama.cpp for CUDA without trouble. If the result holds, a 7B performing like much larger...
A 150-million-parameter model with performance similar to GPT-2 Small has been trained on a single RTX Pro 6000 using 7 billion tokens. Published benchmarks: PIQA 62.24%, Hellaswag 32.20%, Arc-Easy 44.91%, Arc-Challenge 25.00%, Arithmark 3.0 33.90%, ...
An independent researcher has prepared a dense 9.4-billion-parameter model for training on a single card, with distillation from Llama 3 and architectural simplifications. No benchmarks are public, but the signal for on-premise AI is clear: the bottl...
An independent researcher has opened a roughly 9.4B parameter dense model designed for single-card training. The project uses a Llama 3 tokenizer, distilled logit data, and architecture cues from Moonshot and Qwen. The real test is not technical; it ...
In the BabyLM 2026 Strict-Small, Qiushi Engine gained 0.23 Overall points across nine metrics without increasing scale, but by reorganizing experience around contextual dependencies. For teams evaluating self-hosted LLMs, the relevant signal is the d...
Qiushi Engine ran an end-to-end autonomous research program on BabyLM 2026 Strict-Small, using 10 million corpus words and 100 million cumulative word presentations. Three stages linked frontier advancement, principle discovery, and principle-guided ...
A new framework tackles heterogeneous vocabularies in collaborative speculative decoding. X-CoSD limits distribution exchange to the common region between the two models; the extended variant moves resampling to the server and leaves the device with ...
