Pins matching tag:"Apache 2.0"

At its TechXchange event IBM releases Granite 3.0, its third generation of open Apache 2.0 models: 8B and 2B dense language models, Granite Guardian safety models and small mixture-of-experts models.

IBM releases Granite 3.1: every Granite 3 model gains a 128K-token context window, with new Granite Embedding models and function-calling hallucination detection in Granite Guardian 3.1, all under Apache 2.0.

Mistral AI releases Mistral Small 3, a latency-optimized 24B model under Apache 2.0 that it says matches Llama 3.3 70B instruct at more than three times the speed, with over 81% on MMLU.

IBM releases Granite 3.2 with switchable chain-of-thought reasoning in the 8B and 2B Instruct models and its first official vision-language model, Granite Vision 3.2 2B, all under Apache 2.0.

Mistral AI releases Mistral Small 3.1, a 24B Apache 2.0 model with image understanding and a 128k-token context window that Mistral says beats Gemma 3 and GPT-4o Mini at 150 tokens per second.

IBM releases Granite 3.3 8B and 2B Instruct with fill-in-the-middle support and refined reasoning, plus Granite Speech 3.3 8B, its first official speech-to-text model, under Apache 2.0.

IBM releases Granite 4.0 Tiny Preview, an early look at a 7B-parameter hybrid Mamba-2/transformer mixture-of-experts model with 1B active parameters, trained on only 2.5 trillion tokens so far.

Mistral AI releases Mistral Small 3.2, a 24B Apache 2.0 update to Small 3.1 that lifts Arena Hard v2 from 19.56% to 43.1% and cuts infinite generations from 2.11% to 1.29%.

IBM releases Granite 4.0, open Apache 2.0 language models built on a hybrid Mamba-2/transformer architecture that IBM says cuts memory needs by over 70%, the first open models with ISO 42001 certification.

IBM releases Granite 4.0 Nano, its smallest models yet: hybrid and plain-transformer versions of roughly 1B and 350M parameters for edge and on-device use, under Apache 2.0.

Mistral AI releases Mistral 3: Mistral Large 3, a 675B-parameter sparse mixture-of-experts model with 41B active parameters under Apache 2.0, and three dense Ministral 3 models (3B, 8B and 14B).

IBM releases Granite 4.1, dense 3B, 8B and 30B language models under Apache 2.0 whose 8B instruct matches or beats the 32B-parameter Granite 4.0 mixture-of-experts model, alongside new vision, speech, guardian and embedding models.

IBM releases Granite 4.2, three open Apache 2.0 reasoning models (3B, 8B and 30B) with a switchable thinking mode; the 8B scores 47.67 on SWE-bench Verified and 86.67 on AIME25 in IBM's own runs.