Models
Ready To Deploy.
Built to Scale.
No Quantisation, No Distillation.

All Models

Llama 2 7B

Llama 2 7B is the smallest model in Meta’s second-generation LLM series, featuring 7 billion parameters. Built on a decoder-only transformer architecture with rotary position embeddings (RoPE), it supports context lengths of up to 4,096 tokens. It is well-suited for tasks like question answering, reasoning, summarization, and coding.

Phi 3 - 3.8B-Instruct

Phi-3 Mini is a 3.8B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties.

Qwen 2.5 0.5B

Qwen 2.5 0.5B model is the most compact model in the Qwen 2.5 series of large language models developed by Alibaba. It has only 0.5 billion parameters, making it suitable for edge deployment, fast inference, and applications where computational resources are limited. It has a decoder-only transformer architecture with a context length of 32K. lt also supports multilingual tasks.

Qwen 2.5 1.5B

Qwen 2.5 1.5B is a compact large language model from the Alibaba's Qwen 2.5 model series. It is a 1.5 billion parameter decoder-only transformer model, with a context length of 32K. It has a better capability in coding and mathematics, with improvements in instruction following, long texts and structured outputs generation, and understanding structured data. lt also supports multilingual tasks.

Llama 2 13B

Llama 2 13B is the mid-sized model in Meta’s second-generation large language model lineup, containing 13 billion parameters. It uses a decoder-only transformer architecture with rotary position embeddings (RoPE) and supports context lengths up to 4,096 tokens. It is suitable for a range of NLP tasks, including question answering, summarisation, reasoning, and code generation.

Llama 2 Nous Hermes 7B

Nous-Hermes-Llama2-7b is a state-of-the-art language model fine-tuned on over 300,000 instructions. This model was fine-tuned by Nous Research. This Hermes model uses the exact same dataset as Hermes on Llama-1. This model stands out for its long responses, lower hallucination rate, and absence of OpenAI censorship mechanisms. The fine-tuning process was performed with a 4096 sequence length on an 8x a100 80GB DGX machine.

Llama Guard 3 1B

Llama Guard 3-1B is a fine-tuned Llama-3.2-1B pretrained model for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It acts as an LLM – it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated. Llama Guard 3-1B was aligned to safeguard against the MLCommons standardised hazards taxonomy and designed to lower the deployment cost of the moderation system safeguard compared to its predecessors.

Llama Guard 3 8B

Llama Guard 3 is a Llama-3.1-8B pretrained model, fine-tuned for content safety classification. Similar to previous versions, it can be used to classify content in both LLM inputs (prompt classification) and in LLM responses (response classification). It acts as an LLM – it generates text in its output that indicates whether a given prompt or response is safe or unsafe, and if unsafe, it also lists the content categories violated. Llama Guard 3 was aligned to safeguard against the MLCommons standardized hazards taxonomy and designed to support Llama 3.1 capabilities. Specifically, it provides content moderation in 8 languages, and was optimized to support safety and security for search and code interpreter tool calls.

Qwen 2-7B-Instruct

Qwen2 is the new series of Qwen large language models. Qwen2-7B-Instruct supports context lengths up to 131,072 tokens using the YARN technique. The model excels in multilingual language understanding, generation, coding, mathematics, and reasoning. It supports a wide variety of downstream tasks—such as summarization, code generation, creative writing, and Q&A—while remaining fully OpenAI-compatible for seamless integration in inference pipelines

Qwen 2.5-1.5B

Despite its compact size, Qwen2.5‑0.5B shows marked improvements in instruction-following, long-text generation, structured-data understanding (e.g. JSON), and multilingual support across 29+ languages.

Qwen 2.5-3B

Qwen2.5‑3B is a compact, open-weight language model with approximately 3.09 billion parameters. It leverages a decoder-only transformer architecture enhanced with RoPE (rotary embeddings), SwiGLU activations, RMSNorm normalization, and Grouped‑Query Attention across 36 layers. It supports long-context input up to 32,768 tokens and can generate up to 8,192 tokens. Qwen2.5‑3B-instruct is optimized for better instruction-following, structured output, multilingual understanding, and tasks like coding and reasoning, performing strongly across over 29 languages.

Qwen 2.5-14B

Qwen2.5‑14B is a 14.7 billion‑parameter open‑weight model. Built on a decoder-only transformer with advanced features such as RoPE embeddings, SwiGLU activation, RMSNorm, and Grouped Query Attention — organised into 48 layers. It handles context windows up to 128K tokens and can generate up to 8K tokens, with support for over 29 languages. It’s designed for reasoning, coding, long-form generation, and structured output tasks.

Qwen 2.5-32B

Qwen2.5‑32B is a 32.5 billion-parameter open‑weight model. Built with a decoder-only transformer architecture that includes RoPE embeddings, SwiGLU activations, RMSNorm, attention QKV bias, and Grouped‑Query Attention across 64 layers, it supports ultra-long contexts up to 128K tokens and maximum generation lengths of around 8K tokens. It delivers strong multilingual performance, supporting over 29 languages for tasks like reasoning, summarization, coding, and structured-output generation.

Qwen 2.5 Coder Instruct

Qwen2.5‑Coder‑7B‑Instruct is a 7.61 billion-parameter instruction‑tuned model. It builds on the Qwen2.5 architecture—featuring decoder-only transformers with RoPE, SwiGLU activations, RMSNorm normalization, attention QKV bias, and Grouped-Query Attention across 28 layers It supports ultra-long context windows up to 128K tokens and generates outputs up to 8K tokens. Trained on 5.5 trillion tokens of source code, synthetic code/text data, and text-code grounding, it delivers state-of-the-art code generation, reasoning, and bug-fixing performance, while maintaining strong math and general language abilities

Qwen 2.5-7B

Qwen2.5‑7B Instruct is a 7.6 billion‑parameter, instruction‑tuned model. Built on a decoder-only transformer with RoPE, SwiGLU activations, RMSNorm, and Grouped‑Query Attention, it handles extended context windows up to 128K tokens, enabling long-form generation with up to 8K token outputs. This model brings marked improvements in mathematics, coding, instruction-following, and structured output generation. It supports over 29 languages and outperforms similarly sized models in benchmarks like MATH, HumanEval, and Arena-Hard.

Llama 3.2 1B

The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). The Llama 3.2 instruction-tuned text-only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarisation tasks.

Llama 3.2 3B

The Llama 3.2 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction-tuned generative models in 1B and 3B sizes (text in/text out). The Llama 3.2 instruction-tuned text-only models are optimized for multilingual dialogue use cases, including agentic retrieval and summarisation tasks.

Llama 3.1 8B

The Meta Llama 3.1 collection of multilingual large language models (LLMs) is a collection of pretrained and instruction tuned generative model in 8B size (text in/text out). The Llama 3.1 instruction tuned text only model 8B are optimized for multilingual dialogue use cases.