Tag

Qwen3.5-9B-Claude-4.6-HighIQ-THINKING-HERETIC-UNCENSORED

Community fine-tuned multimodal Qwen3.5-9B using Claude 4.6 distilled data to change the model's 'thinking' behavior; offers an uncensored 'heretic' flavor with image-text-to-text I/O, benchmark comparisons, and deployment notes for inference frameworks.

Cosmos3-Nano (NVIDIA)

Generate text, images, video, audio and action/robot trajectories from combined text, image, video, audio and action inputs. A Mixture-of-Transformers omnimodal foundation model (Cosmos3‑Nano, 16B params) focused on Physical AI (robotics, AV, simulation) and optimized for NVIDIA GPU runtimes.

Hermes Agent Reasoning Traces

Provides multi-turn agent trajectories with real tool executions and explicit <think> reasoning blocks for training and evaluating tool-calling agents. Contains two model-sourced configs (Kimi-K2.5, GLM-5.1) totaling ~14.7K samples — useful for SFT, agent-skill research, and tool-integration experiments.

Mistral Medium 3.5 128B

A dense 128B multimodal model with a 256k context window, configurable reasoning effort, and native function-calling for agentic workflows. Supports text+image input, multilingual output, and is released on Hugging Face under a Modified MIT license with revenue-based exceptions.

VoxCPM2

Generates 48kHz multilingual speech from text using a tokenizer-free diffusion-autoregressive TTS architecture, supporting natural-language voice design, controllable cloning, and low-latency streaming. Notable for a 2B-parameter backbone and built-in AudioVAE super-resolution (16k→48k).

GLM-5.1

Generates and iterates on long‑horizon agentic plans and code — designed to stay productive across many rounds of tool calls and experiments. Emphasizes iterative reasoning, stronger repo/terminal automation and code generation than GLM‑5, and can be served locally for research and autonomous-agent workloads.

MiniMax-M2.7

Text-generation LLM designed for agentic workflows: supports multi-agent 'Agent Teams', skill stacks and model self-evolution. Ships on Hugging Face with deployment guides (vLLM, Transformers, SGLang) and is positioned for engineering, tool-calling and productivity use cases.

Hy3 preview

A Mixture-of-Experts instruct-capable LLM (295B total, 21B active) designed for long-context reasoning, code/agent workflows and instruction-following; released by Tencent Hy Team with safetensors weights on Hugging Face.

Kimi K2.6

Multimodal agent model for long-horizon coding, image-text understanding, and autonomous task orchestration. Built as a 1T-parameter Mixture-of-Experts with 256K context and native int4 quantization — intended for coding-driven design, persistent background agents, and swarm-style sub-agent workflows.

Qwen3.6-35B-A3B

Open-weight multimodal 35B Qwen3.6 model in Hugging Face Transformers format that supports image/video/text inputs and native long contexts (262,144 tokens). Emphasizes agentic coding and preserved reasoning traces (thinking), uses an MoE-backed architecture and is designed for self-hosting with vLLM/SGLang/KTransformers; requires multi-GPU resources for production.

SWE-ZERO 12M Trajectories

Provides ~12.29M execution‑free agentic coding trajectories (≈112B tokens) sampled from 122K GitHub PRs to mid‑train code and agent models. Uses bash-only actions (grep, git, sed, etc.) so it scales without Docker; trajectories are unverified and intended for mid-training rather than final SFT.

Qwen3.6-35B-A3B-DFlash

Drafts multiple tokens in parallel with a lightweight block-diffusion drafter to enable speculative decoding for faster LLM inference. Designed to pair with Qwen3.6-35B-A3B and reports up to ~2.9× throughput improvements on common benchmarks.