AI models released in November 2024
21 releases from 13 labs in November 2024, including QwQ-32B-Preview, OLMo 2 and SmolVLM. Every entry is dated to the lab's official announcement and links to it.
- Releases
- 21
- Labs
- 13
- Open weights
- 14
November 2024
21 releasesQwQ-32B-Preview
QwQ-32B-Preview (Qwen with Questions) is an experimental open-weights reasoning model from the Qwen team, focused on advancing AI reasoning in math, code, and general knowledge.
- Modalities
- text → text
- Context
- 32K context
- Parameters
- 32B
- License
- Apache-2.0
- At launch
- Open weights
- Full 32,768-token context length
- Experimental release with limitations such as language mixing and recursive reasoning loops
- Weights under the Apache 2.0 license
OLMo 2
Second generation of fully open OLMo models, trained on up to 5 trillion tokens and released with weights, data, code, recipes, checkpoints and instruct versions.
- Modalities
- text → text
- Parameters
- 7B and 13B
- License
- Apache-2.0
- At launch
- Open weights
- Released in 7B and 13B sizes
- OLMo 2-Instruct models apply the Tülu 3 post-training recipe
- Released under Apache 2.0
SmolVLM
Compact 2B vision-language model built for small memory footprints, released with checkpoints, datasets, training recipes and tools under Apache 2.0.
- Modalities
- text, image → text
- Parameters
- 2B
- License
- Apache-2.0
- At launch
- Open weights
- Released as SmolVLM-Base, SmolVLM-Synthetic and SmolVLM-Instruct
- Trained on the Cauldron and Docmatix datasets
- Integrated into transformers with a fine-tuning script
Stable Diffusion 3.5 Large ControlNets
Blur, Canny and Depth ControlNet models for Stable Diffusion 3.5 Large that add structural control to generations, including tiled upscaling to 8K and 16K resolution.
- Modalities
- text, image → image
- License
- Stability AI Community License
- At launch
- Open weights
- Three control types: Blur, Canny and Depth
- Weights on Hugging Face with ComfyUI support
- Stability AI Community License
Frames
Runway image generation base model focused on stylistic control and visual fidelity, able to hold a specific look across variations; it was later folded into Gen-4 Images.
- Modalities
- text → image
- At launch
- Consumer app, API
- Became available to Unlimited and Enterprise plans on January 17, 2025
- Rolled out inside Gen-3 Alpha and the Runway API
- Frames is now Gen-4 Images
FLUX.1 Tools (Fill, Depth, Canny, Redux)
Suite of control and editing models for FLUX.1: Fill for inpainting and outpainting, Depth and Canny for structural guidance, and Redux for image variation, released as open [dev] weights and [pro] API endpoints.
- Modalities
- text, image → image
- License
- FLUX.1 [dev] Non-Commercial License
- At launch
- Open weights, API, Cloud platforms
- [dev] variants on Hugging Face under the FLUX.1 [dev] Non-Commercial License
- [pro] variants offered in the BFL API
- Depth and Canny were later deprecated in the BFL API; Fill and Redux remain
LTX-Video 0.9
Initial open release of Lightricks' LTX-Video, a diffusion transformer video model supporting text-to-video and image-to-video generation, with a research paper following in December 2024.
- Modalities
- text, image → video
- Parameters
- 2B
- License
- RAIL-M
- At launch
- Open weights
- Text-to-video and image-to-video
- Version 0.9.1 checkpoint with improved quality released December 20, 2024
- Weights on Hugging Face
Tülu 3
Family of post-trained Llama 3.1 models released with all of their data, data mixes, recipes, code, infrastructure and evaluation framework.
- Modalities
- text → text
- Parameters
- 8B and 70B
- License
- Llama 3.1 Community License
- At launch
- Open weights
- Opens up the full post-training recipe
- Built on Llama 3.1 base models
- Later scaled to 405B parameters
DeepSeek-R1-Lite-Preview
DeepSeek-R1-Lite-Preview is a reasoning model preview that shows its thought process in real time. It launched on chat.deepseek.com, with DeepSeek saying open-source models and an API would follow.
- Modalities
- text → text
- At launch
- Consumer app
- Available on DeepSeek's web chat at launch, without API access
- Showed steadily improving results as thought length increased, per DeepSeek
- Preceded the open-weights DeepSeek-R1 release in January 2025
Suno v4
Music model refining v3 with higher audio quality, launched with Remaster for upgrading older tracks, the ReMi lyrics model and v4-powered Covers and Personas.
- Modalities
- text, audio → audio
- At launch
- Consumer app
- Remaster upgrades older songs to v4 quality
- New ReMi lyrics assist model
- Powers Covers and Personas
Mistral Large 24.11
Update of Mistral Large 2 with better long-context understanding, a new system prompt and more accurate function calling, released with weights under the Mistral Research License.
- Modalities
- text → text
- Context
- 125K context
- Parameters
- 123B
- License
- Mistral Research License
- At launch
- Open weights, API, Cloud platforms
- Also called Mistral Large 2.1 in Mistral's changelog
- API model ID mistral-large-2411
- Announced for Google Cloud and Microsoft Azure
Pixtral Large
Frontier-class multimodal model that extends Mistral Large 2 with a 1B-parameter vision encoder, powering document and image understanding in Le Chat.
- Modalities
- text, image → text
- Context
- 125K context
- Parameters
- 124B: 123B decoder + 1B vision encoder
- License
- Mistral Research License
- At launch
- Open weights, API, Consumer app
- 128K-token context fits at least 30 high-resolution images
- API model ID pixtral-large-latest
- Weights released under the Mistral Research License
Qwen2.5-Turbo
Qwen2.5-Turbo is a version of Qwen2.5 that supports a 1M-token context, offered through the Alibaba Cloud Model Studio API. It uses sparse attention mechanisms to speed up long-context processing.
- Modalities
- text → text
- Context
- 1M context
- At launch
- API
- Time to first token for 1M-token context cut from 4.9 minutes to 68 seconds
- Compatible with the OpenAI API format
- Demos available on Hugging Face and ModelScope
JanusFlow
JanusFlow is a unified model for image understanding and generation that integrates an autoregressive language model with rectified flow. DeepSeek reports that rectified flow can be trained within the language model framework without complex architectural changes.
- Modalities
- text, image → text, image
- Parameters
- 1.3B
- License
- DeepSeek Model License
- At launch
- Open weights
- Released as JanusFlow-1.3B on Hugging Face
- Part of DeepSeek's Janus series of unified multimodal models
- Model use under the DeepSeek Model License
Qwen2.5-Coder (0.5B to 32B)
The Qwen2.5-Coder series expanded to six sizes with the release of 0.5B, 3B, 14B, and 32B models, led by Qwen2.5-Coder-32B-Instruct. Qwen says the 32B model matches GPT-4o's coding capabilities.
- Modalities
- text → text
- Context
- 128K context
- Parameters
- 0.5B, 1.5B, 3B, 7B, 14B, 32B
- License
- Apache-2.0 (3B uses the Qwen Research License Agreement)
- At launch
- Open weights
- Adds four sizes to the earlier 1.5B and 7B models
- Full 131,072-token context for the larger models
- Most sizes under Apache 2.0
CogVideoX1.5
CogVideoX1.5 is an upgraded version of Z.ai's open video generation model CogVideoX. The CogVideoX1.5-5B series generates 10-second videos at higher resolution, and its image-to-video variant supports any resolution.
- Modalities
- text, image → video
- Parameters
- 5B
- License
- CogVideoX License
- At launch
- Open weights
- Released as CogVideoX1.5-5B and CogVideoX1.5-5B-I2V
- Diffusers version followed on November 15, 2024
- Weights under the CogVideoX License
Mistral Moderation
LLM-based content moderation classifier that sorts text into nine policy categories, offered as an API with endpoints for raw text and conversational content.
- Modalities
- text → text
- Context
- 8K context
- At launch
- API
- The same model powers moderation in Le Chat
- Covers categories including unqualified advice and PII
- API model ID mistral-moderation-2411
FLUX1.1 [pro] Ultra and Raw modes
High-resolution and photorealism modes for FLUX1.1 [pro]: Ultra generates images at four times standard resolution, up to 4MP, while Raw produces a less synthetic, candid-photography look.
- Modalities
- text → image
- At launch
- API
- Ultra mode supports up to 4MP output
- Raw mode increases diversity in human subjects
- Available through the BFL API
Hunyuan-Large
Hunyuan-Large (Hunyuan-MoE-A52B) is Tencent's open Transformer-based mixture-of-experts language model with 389B total and 52B active parameters, released with a technical report and training and inference tooling.
- Modalities
- text → text
- Context
- 125K context
- Parameters
- MoE: 389B total, 52B active
- License
- Tencent Hunyuan Community License Agreement
- At launch
- Open weights, Cloud platforms
- Pretrained model supports 256K tokens; the Instruct model supports 128K
- Pretrain, Instruct, and Instruct-FP8 checkpoints on Hugging Face
- Integrated into Tencent Cloud's TI platform at launch
Hunyuan3D-1.0
Hunyuan3D-1.0 is a two-stage open framework for text-to-3D and image-to-3D generation that combines fast multi-view diffusion with feed-forward reconstruction, released in lite and standard versions.
- Modalities
- text, image → 3d
- License
- Tencent Hunyuan Non-Commercial License Agreement
- At launch
- Open weights
- Lite model produces a mesh in about 10 seconds and the standard model in about 25 seconds
- Uses Hunyuan-DiT for text-conditioned generation
grok-beta
grok-beta was the preview model that opened xAI's API public beta, a provisional name for a new Grok model then in the final stages of development, supporting function calling and system prompts.
- Modalities
- text → text
- Context
- 125K context
- At launch
- API
- 128,000-token context window
- REST API compatible with OpenAI and Anthropic SDKs
- An image-input preview, grok-vision-beta, followed
Cite and reuse this data
These release records are free to reuse under CC BY 4.0, including commercially. Credit The AI Engineering and link to this page's URL.
Dates are the labs' official announcement dates, and every record links to its source. List fields in the CSV are separated by " | ".
Citation
The AI Engineering Editors. "AI models released in November 2024." The AI Engineering, updated Nov 28, 2024. https://theaiengineering.com/releases/2024/11
HTML link
<a href="https://theaiengineering.com/releases/2024/11">AI models released in November 2024</a> (The AI Engineering)- Full dataset
- 1,167 releases from 55 labs
- Coverage
- to
- Updated
- License
- CC BY 4.0
- Publisher
- The AI Engineering