All local LLM models

Qwen3.5-397B-A17B is a mixture-of-experts model with 397B total parameters and 17B active parameters in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 1M context window, Apache-2.0 license, and local runtimes such as llama.cpp, vllm, sglang.

MiniMax-M2.5 is a mixture-of-experts model with 230B total parameters and 10B active parameters in the MiniMax family. ToolHalla tracks it for chat, coding, research, agentic with a 1.048576M context window, Apache-2.0 license, and local runtimes such as llama.cpp, vllm, sglang.

DeepSeek-R1-671B is a mixture-of-experts model with 671B total parameters and 37B active parameters in the DeepSeek family. ToolHalla tracks it for chat, coding, research, reasoning with a 131k context window, MIT license, and local runtimes such as llama.cpp, vllm, sglang.

Qwen3.5-122B-A10B is a mixture-of-experts model with 122B total parameters and 10B active parameters in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 1M context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Kimi-K2.5 is a mixture-of-experts model with 1T total parameters and 32B active parameters in the Kimi family. ToolHalla tracks it for chat, coding, research, agentic with a 131k context window, MIT (modified) license, and local runtimes such as llama.cpp, vllm, sglang.

GLM-5 is a mixture-of-experts model with 744B total parameters and unknown active parameters in the GLM family. ToolHalla tracks it for chat, coding, research, agentic with a 131k context window, MIT license, and local runtimes such as llama.cpp, vllm, sglang.

Qwen3-235B-A22B is a mixture-of-experts model with 235B total parameters and 22B active parameters in the Qwen family. ToolHalla tracks it for chat, coding, research, reasoning with a 131k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.3-70B-Instruct is a dense 70B parameter model in the Llama family. ToolHalla tracks it for chat, coding, research, creative with a 131k context window, Llama 3.3 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-72B-Instruct is a dense 72B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.1-70B-Instruct is a dense 70B parameter model in the Llama family. ToolHalla tracks it for chat, research, creative, math with a 131k context window, Llama 3.1 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Nous-Hermes-2-Mixtral-8x7B-DPO is a dense 46.7B MoE parameter model in the Nous family. ToolHalla tracks it for chat, creative, coding with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-Coder-32B-Instruct is a dense 32B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, math, research with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DeepSeek-R1-Distill-Llama-70B is a dense 70B parameter model in the DeepSeek family. ToolHalla tracks it for math, research, coding, chat with a 33k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen3-32B is a dense 32B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen3.5-27B is a dense 27B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 1M context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DeepSeek-R1-Distill-Qwen-32B is a dense 32B parameter model in the DeepSeek family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-4-Maverick-17B is a mixture-of-experts model with 400B total parameters and 17B active parameters in the Llama family. ToolHalla tracks it for chat, coding, vision, research with a 1.048576M context window, Llama 4 Community license, and local runtimes such as llama.cpp, vllm, sglang.

Qwen3.5-35B-A3B is a mixture-of-experts model with 35B total parameters and 3B active parameters in the Qwen family. ToolHalla tracks it for chat, coding, agentic, research with a 1M context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-32B-Instruct is a dense 32B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-3-27B is a dense 27B parameter model in the Gemma family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Phi-4-14B-Instruct is a dense 14B parameter model in the Phi family. ToolHalla tracks it for coding, math, research, chat with a 128k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Mistral-Small-24B-Instruct is a dense 24B parameter model in the Mistral family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

InternLM2.5-20B-Chat is a dense 20B parameter model in the InternLM family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Phi-3-medium-128k-instruct is a dense 14B parameter model in the Phi family. ToolHalla tracks it for coding, chat, research, math with a 131k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Mistral-Nemo-12B-Instruct is a dense 12B parameter model in the Mistral family. ToolHalla tracks it for chat, coding, creative with a 128k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen3-30B-A3B is a dense 30B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-4-26B-A4B is a mixture-of-experts model with 25.2B total parameters and 3.8B active parameters in the Gemma family. ToolHalla tracks it for chat, coding, research, math with a 262k context window, Apache-2.0 license, and local runtimes such as huggingface, transformers.

Qwen3.5-9B is a mixture-of-experts model with 9B total parameters in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 262k context window, Apache-2.0 license, and local runtimes such as huggingface, transformers, vllm.

DeepSeek-R1-Distill-Llama-8B is a dense 8B parameter model in the DeepSeek family. ToolHalla tracks it for math, research, chat with a 33k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Mixtral-8x22B-Instruct is a dense 141B MoE parameter model in the Mistral family. ToolHalla tracks it for chat, coding, research, creative with a 66k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-4-Scout-17B is a mixture-of-experts model with 109B total parameters and 17B active parameters in the Llama family. ToolHalla tracks it for chat, coding, vision with a 524k context window, Llama 4 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-Coder-14B-Instruct is a dense 14B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

WizardLM-2-8x22B is a dense 141B MoE parameter model in the WizardLM family. ToolHalla tracks it for chat, coding, research, creative with a 66k context window, Llama 2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-2-27B-Instruct is a dense 27B parameter model in the Gemma family. ToolHalla tracks it for chat, coding, research, creative with a 8k context window, Gemma Terms license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-14B-Instruct is a dense 14B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-3-12B is a dense 12B parameter model in the Gemma family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.2-11B-Vision-Instruct is a dense 11B parameter model in the Llama family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen3.5-4B is a mixture-of-experts model with 4B total parameters in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 262k context window, Apache-2.0 license, and local runtimes such as huggingface, transformers, vllm.

DeepSeek-R1-Distill-Qwen-14B is a dense 14B parameter model in the DeepSeek family. ToolHalla tracks it for math, research, coding with a 33k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Solar-10.7B-Instruct is a dense 10.7B parameter model in the Solar family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DBRX-Instruct is a dense 132B MoE parameter model in the Databricks family. ToolHalla tracks it for chat, coding, research with a 33k context window, Databricks Open license, and local runtimes such as ollama, llama.cpp, vllm.

Ministral-8B-Instruct is a dense 8B parameter model in the Mistral family. ToolHalla tracks it for chat, research with a 128k context window, Mistral Research license, and local runtimes such as ollama, llama.cpp, vllm.

Mixtral-8x7B-Instruct is a dense 46.7B MoE parameter model in the Mistral family. ToolHalla tracks it for chat, coding, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen3-8B is a dense 8B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Command-R-7B is a dense 7B parameter model in the Cohere family. ToolHalla tracks it for research, chat, creative with a 128k context window, CC-BY-NC license, and local runtimes such as ollama, llama.cpp, vllm.

Granite-3.1-8B-Instruct is a dense 8B parameter model in the Granite family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Mistral-7B-Instruct-v0.3 is a dense 7B parameter model in the Mistral family. ToolHalla tracks it for chat, research with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Dolphin-2.9.2-Qwen2-7B is a dense 7B parameter model in the Dolphin family. ToolHalla tracks it for chat, creative, coding with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DeepSeek-Coder-33B-Instruct is a dense 33B parameter model in the DeepSeek family. ToolHalla tracks it for coding, chat, math with a 16k context window, DeepSeek License license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-Coder-7B-Instruct is a dense 7B parameter model in the Qwen family. ToolHalla tracks it for coding, chat with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Falcon-40B-Instruct is a dense 40B parameter model in the Falcon family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

GLM-4.7-9B-Chat is a dense 9B parameter model in the GLM family. ToolHalla tracks it for chat, coding with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.1-8B-Instruct is a dense 8B parameter model in the Llama family. ToolHalla tracks it for chat, research, creative with a 131k context window, Llama 3.1 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-7B-Instruct is a dense 7B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

OpenHermes-2.5-Mistral-7B is a dense 7B parameter model in the Nous family. ToolHalla tracks it for chat, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-2-9B-Instruct is a dense 9B parameter model in the Gemma family. ToolHalla tracks it for chat, research, creative with a 8k context window, Gemma Terms license, and local runtimes such as ollama, llama.cpp, vllm.

Zephyr-7B-beta is a dense 7B parameter model in the Zephyr family. ToolHalla tracks it for chat, creative with a 33k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Vicuna-13B is a dense 13B parameter model in the Vicuna family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Orca-2-13B is a dense 13B parameter model in the Orca family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

CodeLlama-7B-Instruct is a dense 7B parameter model in the CodeLlama family. ToolHalla tracks it for coding with a 16k context window, Llama 2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Neural-Chat-7B-v3.3 is a dense 7B parameter model in the Intel family. ToolHalla tracks it for chat, research with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-4-E4B is a dense 8B parameter model in the Gemma family. ToolHalla tracks it for chat, coding, research, agentic with a 131k context window, Apache-2.0 license, and local runtimes such as huggingface, transformers.

Phi-4-mini-instruct is a dense 3.8B parameter model in the Phi family. ToolHalla tracks it for chat, coding, research with a 128k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Command-R-35B is a dense 35B parameter model in the Cohere family. ToolHalla tracks it for research, chat, coding with a 128k context window, CC-BY-NC license, and local runtimes such as ollama, llama.cpp, vllm.

CodeLlama-34B-Instruct is a dense 34B parameter model in the CodeLlama family. ToolHalla tracks it for coding, math with a 16k context window, Llama 2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

OpenChat-3.6-8B is a dense 8B parameter model in the OpenChat family. ToolHalla tracks it for chat, creative with a 8k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DeepSeek-R1-Distill-Qwen-7B is a dense 7B parameter model in the DeepSeek family. ToolHalla tracks it for math, research, coding with a 33k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-3-4B is a dense 4B parameter model in the Gemma family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Phi-3-mini-4k-instruct is a dense 3.8B parameter model in the Phi family. ToolHalla tracks it for chat, coding with a 4k context window, MIT license, and local runtimes such as ollama, llama.cpp, vllm.

InternLM2.5-7B-Chat is a dense 7B parameter model in the InternLM family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

DeepSeek-Coder-6.7B-Instruct is a dense 6.7B parameter model in the DeepSeek family. ToolHalla tracks it for coding, chat with a 16k context window, DeepSeek License license, and local runtimes such as ollama, llama.cpp, vllm.

StarCoder2-15B-Instruct is a dense 15B parameter model in the StarCoder family. ToolHalla tracks it for coding, math with a 16k context window, OpenRAIL-M license, and local runtimes such as ollama, llama.cpp, vllm.

WizardLM-2-7B is a dense 7B parameter model in the WizardLM family. ToolHalla tracks it for chat, coding with a 33k context window, Llama 2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Vicuna-7B is a dense 7B parameter model in the Vicuna family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Orca-2-7B is a dense 7B parameter model in the Orca family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2-VL-7B-Instruct is a dense 7B parameter model in the Qwen family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

CodeGemma-7B-Instruct is a dense 7B parameter model in the Gemma family. ToolHalla tracks it for coding, chat with a 8k context window, Gemma Terms license, and local runtimes such as ollama, llama.cpp, vllm.

Falcon-7B-Instruct is a dense 7B parameter model in the Falcon family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Qwen2.5-3B-Instruct is a dense 3B parameter model in the Qwen family. ToolHalla tracks it for chat, coding, research with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

CodeLlama-13B-Instruct is a dense 13B parameter model in the CodeLlama family. ToolHalla tracks it for coding, math with a 16k context window, Llama 2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.2-3B-Instruct is a dense 3B parameter model in the Llama family. ToolHalla tracks it for chat, creative with a 131k context window, Llama 3.2 Community license, and local runtimes such as ollama, llama.cpp, vllm.

StarCoder2-7B-Instruct is a dense 7B parameter model in the StarCoder family. ToolHalla tracks it for coding with a 16k context window, OpenRAIL-M license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3.2-1B-Instruct is a dense 1B parameter model in the Llama family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Gemma-2-2B-Instruct is a dense 2B parameter model in the Gemma family. ToolHalla tracks it for chat, creative with a 8k context window, Gemma Terms license, and local runtimes such as ollama, llama.cpp, vllm.

SmolLM2-1.7B-Instruct is a dense 1.7B parameter model in the SmolLM family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

StarCoder2-3B-Instruct is a dense 3B parameter model in the StarCoder family. ToolHalla tracks it for coding with a 16k context window, OpenRAIL-M license, and local runtimes such as ollama, llama.cpp, vllm.

TinyLlama-1.1B is a dense 1.1B parameter model in the TinyLlama family. ToolHalla tracks it for coding, chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Yi-1.5-34B-Chat is a dense 34B parameter model in the Yi family. ToolHalla tracks it for chat, research, creative, math with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Yi-1.5-9B-Chat is a dense 9B parameter model in the Yi family. ToolHalla tracks it for chat, research, creative with a 33k context window, Apache-2.0 license, and local runtimes such as ollama, llama.cpp, vllm.

Llama-3-8B-Instruct is a dense 8B parameter model in the Llama family. ToolHalla tracks it for chat, creative with a 8k context window, Llama 3 Community license, and local runtimes such as ollama, llama.cpp, vllm.