News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
fnMotif-2

Motif-2

A 12.7 billion parameter reasoning model by Motif Technologies.

Motif-2 is a language model developed by Motif Technologies, featuring a 12.7 billion parameter variant. It includes an instruct version and a reasoning-enhanced version designed for complex problem-solving tasks. The model supports long context windows and tool calling capabilities.

Share on WhatsApp

  • Contains 12.7 billion parameters
  • Supports context windows up to 128K tokens
  • Includes a dedicated reasoning-enhanced variant
  • Scores 70 on GPQA-Diamond benchmark
  • Compatible with vLLM for efficient serving

What it is

Motif-2 is a language model family created by Motif Technologies. The primary variant discussed has 12.7 billion parameters. It comes in two main forms: an instruct-tuned version and a reasoning-enhanced version. The reasoning variant is built to improve performance on tasks requiring step-by-step logic. The model is designed to handle long sequences of text and can interact with external tools. It is available for download with open weights.

How it came about

Motif Technologies released the Motif-2 models, including the 12.7 billion parameter instruct and reasoning variants. Detailed information about the development process was noted as pending release at the time of the model card publication. The company provided evaluation results comparing the base instruct model against the reasoning-enhanced version across several benchmarks.

Inside the model

The model uses a decoder-only architecture. It supports a native sequence length of up to 64,000 tokens. Users can extend this context window to 128,000 tokens by applying YaRN scaling techniques. The system supports tool calling through specific parsers like Hermes. It also features a ratio-based thinking budget that guides how many tokens the model uses for internal reasoning before generating a final response. This is managed via custom logits processors.

What it can do

The reasoning-enhanced version shows improved scores on several benchmarks compared to the instruct version. On GPQA-Diamond, it scored 70 compared to 63.6. For LiveBench, it achieved 49.9 against 33.8. LiveCodeBench v5 scores rose from 50.03 to 65. However, GSM8K scores were slightly lower at 95.53 versus 96.13. The model supports chain-of-thought prompting and strict prompt following evaluations.

How to use it

The model weights are available for download via Hugging Face. It is designed to run with vLLM, a library for serving large language models. While official support in vLLM was under review, the developers provided a custom container image for immediate use. Users can run the model locally using command-line tools and configure advanced options for performance.

Questions people ask about Motif-2

What is the difference between the Instruct and Reasoning versions?

The Reasoning version is an enhanced variant of the Instruct model. It is optimized for tasks requiring logical deduction and step-by-step problem solving. Benchmarks show it performs better on GPQA-Diamond and LiveBench than the standard Instruct version.

How long of a context window does Motif-2 support?

The model natively supports a sequence length of up to 64,000 tokens. By using YaRN scaling configurations during deployment, users can extend this context window to support up to 128,000 tokens.

Can Motif-2 use external tools?

Yes, the model supports tool calling. Developers can enable this feature using specific vLLM options such as the Hermes tool-call parser. This allows the model to interact with external functions or APIs during its operation.

Good reading and sources

    Where this page comes from

    There is no English Wikipedia article on Motif-2 yet, so the facts on this page come from the sources listed above: huggingface.co/Motif-Technologies/Motif-2-12.7B-Reasoning. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryLLM wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.

    It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

    The weekly digest

    Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.