Introducing the 𝗪𝗼𝗿𝗹𝗱’𝘀 𝗕𝗲𝘀𝘁 𝗢𝗽𝗲𝗻 𝗦𝗼𝘂𝗿𝗰𝗲 𝟳𝗕 𝗟𝗟𝗠 - OpenChat-3.5-1210, further surpassing ChatGPT and Grok models. This upgrade to the widely adopted OpenChat-3.5 is focused on increasing the performance in one of the most important areas for LLMs - Coding. We achieved a near 15 point increase on HumanEval, while maintaining or improving performance on other benchmarks, making OpenChat-3.5-1210 one of the most capable Generalist models to date. The model is now available on HuggingFace: openchat/openchat-3.5-1210 Try it today with Together AI’s (@TogetherCompute) optimized Inference API: api.together.xyz/playground/cha… We would like to thank RunPod (@RunPod_io) for the support and sponsorship provided for the model!
Additional Benchmarks, including the scores for ChatGPT, Grok, and Open Source LLMs.
@OpenChatDev @runpod_io @togethercompute Interesting variations in LLM performance across different benchmarks. What specific capabilities does Grok-1 have that account for its performance in HumanEval?
@OpenChatDev @runpod_io @togethercompute Dear Santa, All I want for Christmas is for people to stop referring to anything ever as just "ChatGPT" and actually type the 4 extra characters to specify what god damn model they are talking about. 🙏
@OpenChatDev @runpod_io @togethercompute You should always directly link to hugging face (or give a magnet link) in such tweets: huggingface.co/openchat/openc…
@OpenChatDev @runpod_io @togethercompute Can you please differentiate, is this ChatGPT 3.5 or ChatGPT-4
@OpenChatDev @runpod_io @togethercompute Why are the benchmarks only chatgpt and grok?
@OpenChatDev @runpod_io @togethercompute 🔥🔥🔥 Absolute gem of a model! Incredible release y'all!
@OpenChatDev @runpod_io @togethercompute Please pump that MMLU! We need to hug Grok completely!
@OpenChatDev @runpod_io @togethercompute how does this compare with OpenHermes 7B?
@OpenChatDev @runpod_io @togethercompute Are there quantized variants available? Can it run on consumer GPUs?
@OpenChatDev @runpod_io @togethercompute But 7b is too little no? It looses context easily, great progress 💪🏻
@OpenChatDev @runpod_io @togethercompute Legend. Incredible work
@OpenChatDev @runpod_io @togethercompute @AIExplainedYT would appreciate you talk about that on your next video ☺️
@OpenChatDev @runpod_io @togethercompute I like what I am seeing. Sure it trips on gpt-4 level tasks, but so does everything (cloude included). I am really glad to have this in the arsenal!
@OpenChatDev @runpod_io @togethercompute Very cool! Are there any insights against the latest ChatGPT (November)?
@OpenChatDev @runpod_io @togethercompute Is it free from eval data contamination?
@OpenChatDev @runpod_io @togethercompute How does it do in the real world? I kind of disregard evals because it seems gamed & data contamination
@OpenChatDev @runpod_io @togethercompute Would love to try this on @ollama
@OpenChatDev @runpod_io @togethercompute Try out OpenChat-3.5 at openindex.ai/openchat powered by @togethercompute
@OpenChatDev @runpod_io @togethercompute If you are so much better, why you had to copy the name? You hope for people to get confused and dowload you instead? lol
@OpenChatDev @runpod_io @togethercompute How much this model need VRam and have you got any awt (4bit bin) model?
@OpenChatDev @runpod_io @togethercompute Need some tweaks…
