OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Blossomlm

BlossomLM is a family of open-weight general-purpose multimodal language models designed for local deployment, with text and image understanding, reasoning modes and tool use. Its repository documents model variants and inference paths using Transformers, vLLM and llama.cpp.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Blossomlm — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use BlossomLM when you want to deploy an open-weight multimodal model locally and can select a variant and runtime that fit your inference hardware and integration needs.

Who it is for

  • Developers building locally hosted text or image-question-answering applications.
  • Teams integrating tool-using assistants for coding, web development or data visualization.
  • Practitioners choosing an open-weight model to match GPU, CPU or hybrid inference resources.

Documented uses

  • Run a multimodal conversational model locally.
  • Interpret image inputs alongside text.
  • Use reasoning for mathematics, coding and other general assistant tasks.
  • Connect model tool use to agent-style workflows.
Before you choose

Requirements & limits

The repository is public and identifies the project license as Apache-2.0. The README says training data is planned for a future release.

  • The 27B variant is intended for GPU deployments where the weights fit in GPU memory.
  • The 35B-A3B variant is described for CPU or hybrid CPU/GPU inference, including partial CPU offload.
  • The 9B variant is positioned for memory-constrained GPUs, mobile devices and lighter workloads.
  • The repository documents Transformers and vLLM with Safetensors model repositories, and llama.cpp with GGUF repositories.
  • Blossom-V7.1 requires its bundled chat template; multi-turn history should retain complete assistant messages, including reasoning and tool calls.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the 27B model with Transformers

Load Blossom-V7.1-27B for text-only generation using the documented Transformers path.

Before you start

  • PyTorch installed for the target hardware
  • Transformers 5.12.1 or later
  • Accelerate

Steps

  1. Install or upgrade Transformers and Accelerate with the command below.

    Documented command
    pip install -U "transformers>=5.12.1" accelerate

  2. Load the Azure99/Blossom-V7.1-27B tokenizer and model with AutoTokenizer.from_pretrained and AutoModelForCausalLM.from_pretrained, using the bundled model ID, torch.bfloat16, and device_map="auto".

  3. Apply the tokenizer’s chat template to the messages with add_generation_prompt=True, generate a response, and decode the generated tokens.

Expected outcome

A local text-only generation path for the 27B model; this example skips the vision encoder.

Guide limitations

  • This documented path is text-only and does not demonstrate image input.
  • The repository requires the bundled chat template and advises against replacing or combining it with a native Qwen template.
  • The README recommends replaying the complete assistant message, including reasoning and tool calls, in multi-turn history.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
BlossomlmRun a multimodal conversational model locally.Self-hostedThe repository is public and identifies the project license as Apache-2.0. The README says training data is planned for a future release.
Pulse Of MotionEstimate a video's physical frame rate from its visual dynamics.LinuxThe repository is public. Its README describes a released checkpoint that downloads from Hugging Face on first inference, with a manual download option. The strongest internal model is not fully released; the README directs interested users to contact the project about access. No license information is provided in the repository evidence.
Neural Image AssessmentTrain the model using AVA images and rating annotations.Cloud, Linux, macOS, Windows
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
8 October 2026
Original website title
Azure99/BlossomLM · Official homepage title · retained in its source language
How we review sourcesSuggest a correction