DeepSeek flagship 1.6T MoE with 49B active, 1M context, MIT weights. API $0.29/$1.20 cached $0.03.

Alibaba Qwen3.8 2.4T MoE with 95B active, 1M context, multimodal input. Coder $0.14/$0.80; 397B $0.29/$1.20.

Alibaba's open-weight 27B dense companion to Qwen3.8-Max: Apache 2.0 license, multimodal image-text input, built for local deployment and small-batch inference.

Moonshot AI's open-weight 2.8T multimodal agentic model with 1M-token context, the world's first open 3T-class model rivaling closed frontier models.

MiniMax-Music3 is an open-weight music model that creates full songs from prompts and lyrics, with a new vocal engine and official ComfyUI support.

Liquid AI's open-weight 2.6B dense model trained for on-device agentic workloads, with 128K context and native tool calling.

Llama 3.1 Nemotron 70B Instruct is NVIDIA's RLHF-tuned Llama 3.1 70B, reporting Arena Hard 85.0 and MT-Bench 8.98, with 128K input context.

NVIDIA's efficient open-weight 30B MoE hybrid model with 3B active parameters, 1M-token context, and single-GPU deployment for local reasoning and coding.

cogvlm-base-490-hf is an open CogVLM-17B vision-language checkpoint from Z.ai (formerly THUDM) with a visual expert module and Apache-2.0 code.

Gemma 2 27B is Google's open-weights 27B decoder-only text model, trained on 13T tokens, in pretrained and instruction-tuned variants under the Gemma license.

Gemma 2 9B is Google's open-weights 9B text model released June 27, 2024, trained on 8T tokens; Gemma 4 is now the latest Gemma generation.

Euryale v2.2 is Sao10K's Llama 3.1 70B finetune for roleplay and creative writing, trained 2 epochs on about 4 days of 8x A100, license cc-by-nc-4.0.

Llama 3 Euryale 70B v2.1 is Sao10K's roleplay-oriented 70B LoRA fine-tune on Llama 3, released under the non-commercial CC-BY-NC-4.0 license.

Llama 2 70B Chat is Meta's July 2023 open-weights chat model with 70B parameters, a 4K context and 2T training tokens, under the Llama 2 license.

Mistral Nemo 12B (open-mistral-nemo-2407) is an Apache 2.0 model with 128k context; Mistral deprecated it May 22, 2026 and retired it July 31, 2026.

MythoMax-L2-13b is Gryphe's Llama 2 13B merge of MythoLogic-L2 and Huginn for roleplay and storywriting, with an Alpaca-style prompt and a legacy label.

Hermes 3 405B is Nous Research's full-parameter fine-tune of Llama 3.1 405B with function calling, JSON mode and ChatML, described in an August 2024 report.

Hermes 2 Pro Llama-3 8B is NousResearch's open-weights fine-tune of Llama 3 8B with ChatML, function calling and JSON mode, and a Llama 3 license.

Rocinante 12B is TheDrummer's 12B fine-tune for roleplay and adventure storytelling, in v1 and v1.1, with ChatML, Alpaca and Mistral templates.

WizardLM-2 7B is a Microsoft WizardLM team chat model fine-tuned from Mistral-7B-v0.1 under Apache 2.0, announced April 15, 2024 with the 8x22B and 70B models.

WizardLM-2 8x22B is the largest of three open WizardLM-2 models announced April 15, 2024, released under Apache 2.0 by the WizardLM team.

llava-v1.6-34b-hf is the Hugging Face Transformers port of LLaVA-NeXT 34B, an open vision-language model announced January 30, 2024 on Nous-Hermes-2-Yi-34B.

BAAI bge-reranker-v2.5-gemma2-lightweight is a roughly 9B multilingual reranker on Gemma 2 9B with token compression and layer cutoffs.

MN-12B-Inferor-v0.0 is a community 12B merge by Infermatic AI, built with mergekit from Mistral Nemo-based models on a Magnum v4 base.