maternion

LightOnOCR-2

Best OCR model. LightOnOCR-2-1B is our flagship OCR model, refined with RLVR training for maximum accuracy. We recommend this variant for most OCR tasks.

vision tools 1b

14.4K Pulls 2 Tags Updated 8 months ago

fara

Fara-7B is Microsoft's first agentic small language model (SLM) designed specifically for computer use. (Template supports tools)

vision tools 7b

5,607 Pulls 4 Tags Updated 10 months ago

hy-mt2

HY-MT2 1.8B, 7B, and 30B-A3B (MoE) multilingual translation models designed for complex real-world scenarios, it supports translation among 33 languages and effectively follows translation instructions.

1.8b 7b 30b

4,947 Pulls 7 Tags Updated 4 months ago

Qianfan-OCR

Qianfan-OCR is a 4B VL model that replaces multi-step OCR pipelines with a single system, enabling direct image-to-Markdown conversion and advanced document understanding.

vision tools thinking 4b

2,438 Pulls 3 Tags Updated 6 months ago

lfm2

LFM2-8B-A1B is the best on-device MoE in terms of both quality (comparable to 3-4B dense models) and speed (faster than Qwen3-1.7B). Updated LFM2.5-8B-A1B is now available!

tools 8b

2,134 Pulls 4 Tags Updated 6 months ago

lfm2.5

LFM2.5-350M & LFM2.5-8B-A1B are fast, memory-efficient hybrid MoE language models. Unsloth Dynamic (UD) quants available.

tools thinking 350m 8b

1,848 Pulls 15 Tags Updated 3 months ago

spark-x2.5-heretic

Spark-X2.5-4B abliterated (uncensored) - efficient AI model for chat, writing, coding, reasoning, tools, and agents, 1M-token context, 200+ languages, with refusal behavior removed.

4b

1,596 Pulls 5 Tags Updated 1 week ago

minicpm5

MiniCPM5-2B is a dense 2B Transformer for on-device deployment. It achieves 2B-class SOTA, remains competitive with 4B models overall, and excels in coding, math, long-context, tool use, and agentic tasks.

2b

1,520 Pulls 4 Tags Updated 3 weeks ago

ling-3.0-tiny

Ling-3.0-tiny is a lightweight 7.9B-parameter hybrid reasoning MoE with just 1.3B active parameters per token, offering strong reasoning and agentic capabilities at low inference cost.

8b

1,489 Pulls 10 Tags Updated 1 month ago

mimo-v2.6

MiMo-V2.6-Distill-9B (Qwen3.5-9B SFT): coding, agent tasks, visual coding, cybersecurity. Thinking + instruct quants, multimodal.

vision 9b

1,484 Pulls 9 Tags Updated 1 week ago

mai-ui

MAI-UI is a family of large-scale GUI foundation agents (2B,8B) built for real-world, interactive human-computer control across desktop and mobile Uls. (Template supports tools)

vision tools 2b 8b

1,347 Pulls 8 Tags Updated 8 months ago

strand-rust-coder

The model fine-tunes Qwen2.5-Coder-14B for Rust-specific programming tasks using a 191K-example synthetic dataset built via multi-model generation and peer-reviewed validation.

tools 14b

1,328 Pulls 3 Tags Updated 7 months ago

mimo-v2.6-heretic

Uncensored/Abliterated MiMo-V2.6-Distill-Qwen-9B: coding, agent tasks, visual coding, cybersecurity. Thinking+instruct variants.

vision 9b

1,043 Pulls 9 Tags Updated 1 week ago

AgentCPM-Explore

AgentCPM-Explore is a small, open‑source deep research model designed for long‑horizon, complex reasoning tasks on devices. (Template supports tools)

tools thinking 4b

893 Pulls 1 Tag Updated 8 months ago

NuMarkdown-Thinking

It is the first reasoning OCR VLM. It is specifically trained to convert documents into clean Markdown files, well suited for RAG applications.

vision thinking 8b

780 Pulls 2 Tags Updated 8 months ago

spark-x2.5

Spark-X2.5-4B is an efficient, open-source AI model built for chat, writing, coding, reasoning, tools, and agents, supporting 1M-token context and 200+ languages.

4b

678 Pulls 4 Tags Updated 1 week ago

minicpm5-heretic

Uncensored/Abliterated MiniCPM5-2B with only 10/100 refusals. It achieves 2B-class SOTA, remains competitive with 4B models overall.

2b

620 Pulls 5 Tags Updated 1 week ago

manim-coder

Fine-tuned manim model using Qwen2.5-Coder-14B-Instruct trained on 3blue1brown-manim dataset with 2,407 examples.

tools 14b

561 Pulls 3 Tags Updated 7 months ago

sweep-next-edit-1.5B

Sweep Next-Edit a 1.5B parameter model for next-edit autocomplete, quantized to Q8_0 GGUF format. (Template supports tools)

tools

448 Pulls 1 Tag Updated 8 months ago

context-1

Context-1 is a 20B agentic search model that breaks complex queries into steps, retrieves relevant info iteratively, and manages its context to support multi-hop reasoning alongside a larger model.

tools thinking 20b

128 Pulls 3 Tags Updated 6 months ago

neohorse-1

NeoHorse-1-9B is a 9B causal language model and an initial prototype on the path toward recursive self-improvement (RSI).

9b

59 Pulls 4 Tags Updated 1 week ago

Trying to help at https://discord.gg/ollama