From 002791b2e8c622507fabb675652e270cba09e717 Mon Sep 17 00:00:00 2001 From: Akemi Izuko Date: Mon, 18 Nov 2024 20:30:31 -0700 Subject: [PATCH] Llama: update history for llama3 --- src/content/llama/a-history-of-llamas.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/content/llama/a-history-of-llamas.md b/src/content/llama/a-history-of-llamas.md index c040629..c3ee1ce 100644 --- a/src/content/llama/a-history-of-llamas.md +++ b/src/content/llama/a-history-of-llamas.md @@ -1,7 +1,7 @@ --- title: 'A Brief History of Local Llamas' description: 'A Brief History of Local Llamas' -updateDate: 'Feb 18 2024' +updateDate: 'Nov 17 2024' heroImage: '/images/llama/tiny-llama-logo.avif' --- @@ -204,6 +204,8 @@ how much impact they have in a retrospective: - **2024-07-16**: Mistral releases codestral-mamba 7B, with up to 256k "tested" tokens of support. Codestral-mamba 7B is under Apache 2.0, marking the first major foundational model provider not using a transformers llama + - **2024-07-23**: LLaMA 3.1 405B, an open weights model by Meta, finally + definitively defeats GPT-4 across all relevant benchmarks! - **2024-09-12**: OpenAI releases their o1 series. These models use hidden autoregression to "reason" better across tasks. This has been well observed before, but open llamas don't have access to reasoning-style data (expensive)