---
title: "AI models released in February 2024"
type: "release-archive"
canonical_url: "https://theaiengineering.com/releases/2024/02"
markdown_url: "https://theaiengineering.com/releases/2024/02.md"
author: "The AI Engineering Editors"
publisher: "The AI Engineering"
updated: "2024-02-28"
citation: "The AI Engineering Editors. \"AI models released in February 2024.\" The AI Engineering, updated Feb 28, 2024. https://theaiengineering.com/releases/2024/02"
license: "CC BY 4.0"
license_url: "https://creativecommons.org/licenses/by/4.0/"
data_csv: "https://theaiengineering.com/releases.csv"
data_json: "https://theaiengineering.com/releases.json"
---

# AI models released in February 2024

13 releases from 9 labs, 6 with open weights. Dates are the labs' official announcement dates.

## February 2024

### Ideogram 1.0

Ideogram text-to-image model focused on text rendering, photorealism and prompt adherence, released to all users with Magic Prompt to help write detailed prompts.

- Date: 2024-02-28
- Lab: Ideogram
- Type: Image
- Weights: Proprietary
- Modalities: text → image
- Available at launch: Consumer app
- Permalink: https://theaiengineering.com/releases/2024/02#ideogram-1-0

Highlights:

- Paid plans added private generation, image upload and Ideogram Editor
- Magic Prompt expands short prompts into detailed ones

Source: [Ideogram on X: Introducing Ideogram 1.0: the most advanced text-to-image model](https://x.com/ideogram_ai/status/1762881278955700270)

### Mistral Large

Mistral AI's flagship proprietary model at launch, fluent in five European languages with native function calling, offered on la Plateforme and Microsoft Azure.

- Date: 2024-02-26
- Lab: Mistral AI
- Type: LLM
- Weights: Proprietary
- Modalities: text → text
- Context window: 32,000 tokens
- Available at launch: API, Cloud platforms
- Permalink: https://theaiengineering.com/releases/2024/02#mistral-large

Highlights:

- 32K-token context window
- Native function calling and JSON output mode
- Azure was the first distribution partner
- API model ID mistral-large-2402

Source: [Au Large | Mistral AI](https://mistral.ai/news/mistral-large)

### Mistral Small (24.02)

Latency- and cost-optimized proprietary model released alongside Mistral Large, positioned as an intermediary between Mistral's open-weight offering and its flagship, with RAG and function calling support.

- Date: 2024-02-26
- Lab: Mistral AI
- Type: LLM
- Weights: Proprietary
- Modalities: text → text
- Context window: 32,000 tokens
- Available at launch: API
- Permalink: https://theaiengineering.com/releases/2024/02#mistral-small-2402

Highlights:

- Supports function calling and JSON mode
- API model ID mistral-small-2402
- The earlier mistral-small endpoint was renamed open-mixtral-8x7b

Source: [Au Large | Mistral AI](https://mistral.ai/news/mistral-large)

### Stable Diffusion 3

Stable Diffusion 3 suite combining a diffusion transformer architecture with flow matching, ranging from 800M to 8B parameters, announced in early preview with improved multi-subject prompts and spelling.

- Date: 2024-02-22
- Lab: Stability AI
- Type: Image
- Weights: Proprietary
- Modalities: text → image
- Parameters: 800M to 8B
- Available at launch: API
- Permalink: https://theaiengineering.com/releases/2024/02#stable-diffusion-3

Highlights:

- Announced with a waitlist for an early preview
- Stable Diffusion 3 and SD3 Turbo became available in the Stability API on April 17, 2024
- Multimodal Diffusion Transformer uses separate weights for image and language

Source: [Stable Diffusion 3](https://stability.ai/news-updates/stable-diffusion-3)

### Gemma

Gemma is Google's first family of lightweight open models, released in 2B and 7B sizes, each with pre-trained and instruction-tuned variants. Its terms of use permit responsible commercial usage and distribution by organizations of any size.

- Date: 2024-02-21
- Lab: Google
- Type: Small model
- Weights: Open weights
- Modalities: text → text
- Context window: 8,192 tokens
- Parameters: 2B, 7B
- License: Gemma Terms of Use
- Available at launch: Open weights, Cloud platforms
- Permalink: https://theaiengineering.com/releases/2024/02#gemma

Highlights:

- Trained with an 8,192-token context length
- Weights on Kaggle and Hugging Face; deployable on Vertex AI and GKE
- Gemma 1.1 update released April 5, 2024

Source: [Gemma: Introducing new state-of-the-art open models](https://blog.google/technology/developers/gemma-open-models/)

### Gemini 1.5 Pro

Gemini 1.5 Pro was the first model in Google's second Gemini generation, a mid-size multimodal model that performed at a similar level to 1.0 Ultra and introduced an experimental context window of up to 1 million tokens.

- Date: 2024-02-15
- Lab: Google
- Type: LLM
- Weights: Proprietary
- Modalities: text, image, audio, video → text
- Context window: 1,000,000 tokens
- Available at launch: API, Cloud platforms, Consumer app
- Permalink: https://theaiengineering.com/releases/2024/02#gemini-1-5-pro

Highlights:

- Launched with a standard 128,000-token window and up to 1 million tokens for a limited group in private preview
- Public preview in the Gemini API in 180+ countries began Apr 9, 2024, adding native audio understanding
- Stable gemini-1.5-pro-001 listed as generally available in the Gemini API changelog on May 23, 2024

Source: [Introducing Gemini 1.5, Google's next-generation AI model](https://blog.google/innovation-and-ai/products/google-gemini-next-generation-model-february-2024/)

### Aya 101

Open-source massively multilingual research model from Cohere For AI covering 101 languages, released with the largest multilingual instruction fine-tuning dataset at the time.

- Date: 2024-02-13
- Lab: Cohere
- Type: LLM
- Weights: Open weights
- Modalities: text → text
- Parameters: 13B
- License: Apache-2.0
- Available at launch: Open weights
- Permalink: https://theaiengineering.com/releases/2024/02#aya-101

Highlights:

- Built with more than 3,000 researchers from 119 countries
- Released with a 513 million instance dataset covering 114 languages
- Expands coverage to more than 50 previously unserved languages

Source: [C4AI Launches Aya, an LLM Covering More Than 100 Languages](https://cohere.com/blog/aya)

### Reka Flash and Reka Edge

Two multimodal, multilingual models: Reka Flash, a 21B model trained from scratch for fast, high-quality workloads, and the compact 7B Reka Edge for on-device and local use, both in public beta.

- Date: 2024-02-12
- Lab: Reka
- Type: Multimodal
- Weights: Proprietary
- Modalities: text, image, video → text
- Parameters: Flash: 21B; Edge: 7B
- Available at launch: Consumer app
- Permalink: https://theaiengineering.com/releases/2024/02#reka-flash

Highlights:

- Pretrained on text from 32 languages
- Image and video understanding, including video captioning and video question answering
- Instruction tuned, then trained with RLHF (PPO) using Reka Flash as the reward model
- Public beta in the Reka Playground at launch

Source: [Reka Flash: Efficient and Capable Multimodal Language Models](https://reka.ai/news/reka-flash-efficient-and-capable-multimodal-language-models)

### Stable Cascade

Research-preview text-to-image model built on the Würstchen architecture, using a three-stage design that decouples text-conditional generation from decoding to make fine-tuning more efficient.

- Date: 2024-02-12
- Lab: Stability AI
- Type: Image
- Weights: Open weights
- Modalities: text → image
- Parameters: Stage C: 1B or 3.6B
- License: Stability AI Non-Commercial Research Community License
- Available at launch: Open weights
- Permalink: https://theaiengineering.com/releases/2024/02#stable-cascade

Highlights:

- Stage C released in 1B and 3.6B parameter versions
- Non-commercial license
- Training and inference code on GitHub

Source: [Introducing Stable Cascade](https://stability.ai/news-updates/introducing-stable-cascade)

### Gemini 1.0 Ultra

Gemini 1.0 Ultra was the largest and most capable model in Google's first Gemini generation. It reached the public through Gemini Advanced on the day Bard was renamed Gemini.

- Date: 2024-02-08
- Lab: Google
- Type: LLM
- Weights: Proprietary
- Modalities: text, image → text
- Available at launch: API, Cloud platforms, Consumer app
- Permalink: https://theaiengineering.com/releases/2024/02#gemini-1-0-ultra

Highlights:

- Announced Dec 6, 2023 as the largest of the three Gemini 1.0 sizes (Ultra, Pro and Nano)
- Launched in Gemini Advanced on Feb 8, 2024, alongside a new Gemini mobile app
- Developers and Cloud customers could begin building with 1.0 Ultra through the Gemini API and Vertex AI from Feb 15, 2024

Source: [Google Bard is now Gemini: How to try Ultra 1.0 and new mobile app](https://blog.google/products-and-platforms/products/gemini/bard-gemini-advanced-app/)

### DeepSeekMath 7B

DeepSeekMath 7B is a math-specialized model initialized from DeepSeek-Coder-v1.5 7B and further pre-trained on 500B tokens of math-related, natural language, and code data. Its RL variant was trained with DeepSeek's Group Relative Policy Optimization (GRPO).

- Date: 2024-02-05
- Lab: DeepSeek
- Type: Reasoning
- Weights: Open weights
- Modalities: text → text
- Parameters: 7B
- License: DeepSeek Model License
- Available at launch: Open weights
- Permalink: https://theaiengineering.com/releases/2024/02#deepseekmath-7b

Highlights:

- Released as Base, Instruct, and RL models
- Introduced the GRPO reinforcement learning algorithm
- Supports commercial use under the DeepSeek Model License

Source: [deepseek-ai/deepseek-math-7b-rl (Hugging Face)](https://huggingface.co/deepseek-ai/deepseek-math-7b-rl)

### Qwen1.5

Qwen1.5 is the next iteration of the Qwen series, open-sourcing base and chat models across multiple sizes with quantized versions. It improved the developer experience with Hugging Face Transformers integration.

- Date: 2024-02-04
- Lab: Alibaba Qwen
- Type: LLM
- Weights: Open weights
- Modalities: text → text
- Context window: 32,768 tokens
- Parameters: 0.5B, 1.8B, 4B, 7B, 14B, 72B
- License: Tongyi Qianwen License Agreement
- Available at launch: Open weights
- Permalink: https://theaiengineering.com/releases/2024/02#qwen1-5

Highlights:

- All models support a context length of up to 32,768 tokens
- Released just before the Chinese New Year with GPTQ, AWQ, and GGUF quantized models; code merged into Hugging Face Transformers
- Weights under the Tongyi Qianwen License Agreement

Source: [Introducing Qwen1.5](https://qwenlm.github.io/blog/qwen1.5/)

### OLMo

Ai2's first truly open language model, released with its full pretraining data, training code, model weights and evaluation tools so researchers can study how it was built.

- Date: 2024-02-01
- Lab: Ai2
- Type: LLM
- Weights: Open weights
- Modalities: text → text
- Context window: 2,048 tokens
- Parameters: 1B and 7B
- License: Apache-2.0
- Available at launch: Open weights
- Permalink: https://theaiengineering.com/releases/2024/02#olmo

Highlights:

- Pretrained on the open Dolma corpus of about three trillion tokens
- Built with partners including the Kempner Institute, AMD and the LUMI supercomputer
- Available on Hugging Face and GitHub

Source: [Hello OLMo: A truly open LLM | Ai2](https://allenai.org/blog/hello-olmo-a-truly-open-llm-43f7e7359222)

## 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 " | ".

The AI Engineering Editors. "AI models released in February 2024." The AI Engineering, updated Feb 28, 2024. https://theaiengineering.com/releases/2024/02

HTML link:

```html
<a href="https://theaiengineering.com/releases/2024/02">AI models released in February 2024</a> (The AI Engineering)
```

- Title: AI models released in February 2024
- Author: The AI Engineering Editors
- Publisher: The AI Engineering
- Updated: 2024-02-28
- URL: https://theaiengineering.com/releases/2024/02
- Markdown: https://theaiengineering.com/releases/2024/02.md
- License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
- Full dataset as CSV: https://theaiengineering.com/releases.csv
- Full dataset as JSON: https://theaiengineering.com/releases.json
