Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
🚀🔥 Don't miss upstage/solar-pro-preview-instruct on the Open LLM Leaderboard! • First 22B model competing with 70B models 🦾 • Truly impressive scores – 39.61 average, 84.16 IFEval, 54.82 BBH, and 20.09 for MATH! Check out the full breakdown in the screenshot! 📊 • MIT… https://t.co/Yp2iewhZEA
Across 30 tasks, @UpstageAI's Solar Pro Preview fine-tuned outperforms every top fine-tuned open-source #LLM and #GPT4 and #GPT4o-mini base models! See more at our new Fine-Tuning Leaderboard: https://t.co/Gx91z5mUS0 Try #SolarProPreview for free: https://t.co/faiEqG7l8p https://t.co/aYhQyNBP38
New models on the EQ-Bench leaderboards: upstage/solar-pro-preview-instruct ifable/gemma-2-Ifable-9B (new leader!) Sao10K/MN-12B-Lyra-v1 FluffyKaeloky/Luminum-v0.1-123B c4ai-command-r-08-2024 arcee-ai/Arcee-Scribe TheDrummer/Theia-21B-v1 ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.1 https://t.co/UsHo05zsHy
UpstageAI's Solar Pro Preview has achieved the top position under 70B models on the HuggingFace OpenLLM Leaderboard. The model, which has 22 billion parameters, outperforms other top fine-tuned open-source LLMs, including GPT-4 and Llama-3.1-8b-instruct, across 30 tasks and 24 tasks. It fits on a single GPU, making it cost-efficient for serving. The model has impressive scores, including a 39.61 average, 84.16 IFEval, 54.82 BBH, and 20.09 for MATH. Solar Pro Preview is now available on VESSLAI, following a collaboration between VESSLAI and UpstageAI at OCW2024.
