Microsoft launches Phi-3-mini
On 23 April 2024 Microsoft introduced the Phi-3 family of small language models, starting with Phi-3-mini, released on Azure AI Studio, Hugging Face and Ollama.
Total 671 · Clear filters
Apple releases the OpenELM open language models
On 22 April 2024 Apple researchers posted the OpenELM paper to arXiv and released eight small language models, with training logs, checkpoints, configurations and code for running the models on Apple devices.
Meta releases Llama 3
On 18 April 2024 Meta released Llama 3 in 8B and 70B sizes, both pretrained and instruction-tuned. The models were pretrained on over 15 trillion tokens and made available through major cloud providers and Hugging Face under the Meta Llama 3 Community License.
MiniMax launches the abab 6.5 mixture-of-experts series
On 17 April 2024, MiniMax released the abab 6.5 series, a family of mixture-of-experts language models reported to support a 200,000-token context. It extended the company's proprietary abab line, which began with abab1 in 2022.
Mistral AI releases Mixtral 8x22B
On 17 April 2024 Mistral AI announced Mixtral 8x22B, a sparse mixture-of-experts language model with 141 billion total but only 39 billion active parameters, released under Apache 2.0 with a 64K context window.
Boston Dynamics retires hydraulic Atlas and unveils electric Atlas
In April 2024 Boston Dynamics retired its hydraulic Atlas and introduced a fully electric humanoid, promising more strength, dexterity and range of motion, with Hyundai's factories as the first testing ground.
Cohere launches Command R+, first on Microsoft Azure
On 4 April 2024 Cohere introduced Command R+, a retrieval-augmented model for enterprise workloads. It was first offered on Microsoft Azure, where Microsoft said it works with Azure AI Search and Azure AI Content Safety.
Stability AI releases Stable Audio 2.0
On 3 April 2024 Stability AI released Stable Audio 2.0, a model that generates stereo tracks of up to three minutes, adds audio-to-audio transformation, and was trained on more than 800,000 licensed files.
Langboat open-sources the Mengzi3 model family
On 29 March 2024 Langboat released Mengzi3, an open-weight language model family in 8B and 13B sizes under Apache-2.0, with base weights first and a chat version following later.
AI21 Labs releases Jamba, a hybrid Mamba-Transformer model
On 28 March 2024 AI21 Labs released Jamba, an open-weights model that combines Mamba state-space layers with Transformer attention. It has 52 billion total parameters and a 256k-token context.
xAI announces Grok-1.5 with a 128K context
On 28 March 2024 xAI announced Grok-1.5, with better reasoning and a 128,000-token context window. About two weeks later it previewed Grok-1.5V, its first multimodal model.
Databricks releases the 132B mixture-of-experts model DBRX
Databricks released DBRX, a mixture-of-experts model shipped in base and instruct versions, on 27 March 2024. The company framed it as competition for Mixtral and Grok-1 rather than closed frontier systems, and reported it ahead of LLaMA2-70B Chat on every benchmark it published.
StepFun debuts with the Step-1, Step-1V and Step-2 preview models
On 23 March 2024 StepFun made its public debut in Shanghai, announcing Step-1, Step-1V and a preview of the trillion-parameter Step-2, plus two consumer apps.
Sakana AI merges open models by evolutionary search to build a Japanese LLM
On 19 March 2024, Sakana AI posted a paper on using evolutionary search to merge open models without extra training data or compute, and released a 7B Japanese model that also handles mathematics as an experiment.
NVIDIA announces the Blackwell platform
On 18 March 2024 NVIDIA announced Blackwell, the GPU platform succeeding Hopper, claiming it can run real-time generative AI on trillion-parameter models at up to 25 times lower cost and energy.
NVIDIA announces Project GR00T
In March 2024 NVIDIA announced Project GR00T, a general-purpose foundation model for humanoid robots. A year later it released Isaac GR00T N1, which it presented as a fully customizable humanoid model.
xAI releases Grok-1's weights under Apache 2.0
On 17 March 2024 xAI released the base weights and architecture of Grok-1, a 314-billion-parameter mixture-of-experts model, under the Apache 2.0 license.
Claude 3 Haiku becomes available
On 13 March 2024 Anthropic made Claude 3 Haiku available, calling it the fastest and most affordable model in its intelligence class. It was priced at a one-to-five input-to-output ratio and accepted image input.
Cohere releases Command R for enterprise RAG and tool use
On 11 March 2024 Cohere released Command R, a 35-billion-parameter model built for retrieval-augmented generation and tool use. Its weights were released for research use on Hugging Face.
Anthropic introduces the Claude 3 model family
On 4 March 2024 Anthropic announced the Claude 3 family: Haiku, Sonnet and Opus, in ascending order of capability. All three added vision and launched with a 200K-token context window.
BigCode and NVIDIA release StarCoder2 and The Stack v2
On 28 February 2024 BigCode and NVIDIA released StarCoder2 in 3B, 7B and 15B sizes. The new training set, The Stack v2, is 67.5 TB and is built from the Software Heritage archive.