GPT-4.5 will have a much better accuracy. Accuracy jumps from 28% (4o) to 78% (4.5), with o1 being at 55%
@testingcatalog
-
GPT-4.5 System Card Details
By
–
What we know so far A GPT-4.5 system card https://
x.com/scaling01/stat
us/1895180769171005464
… -
GPT-4.5 system card and SWE-bench eval
By
–
GPT-4.5 system card and SWE-bench eval, in particular Looks like we will be getting a minor upgrade today https://
x.com/scaling01/stat
/scaling01/status/1895180769171005464
… -
Gemini iOS adds recall feature for past chats
By
–
Gemini app for iOS now supports “recall” feature – it can refer to past conversations and respond with links to previous chats.
-
Ideogram 2a: Faster, Cheaper Text-to-Image Model
By
–
Ideogram released "Ideogram 2a" – a new version of their text-to-image model which is faster and more affordable 🔥 https://t.co/YUZ6jV7erl pic.twitter.com/KOKiScm986
— 🚨 AI News | TestingCatalog (@testingcatalog) 27 février 2025Ideogram released "Ideogram 2a" – a new version of their text-to-image model which is faster and more affordable
-
Hume’s Octave: TTS with Understanding
By
–
Hume released a big Octave upgrade – the first text-to-speech model that understands what it’s saying.
— 🚨 AI News | TestingCatalog (@testingcatalog) 26 février 2025
Some voice samples below 👇 https://t.co/63cpMn8Ymc pic.twitter.com/HV1KxQrcCLHume released a big Octave upgrade – the first text-to-speech model that understands what it’s saying. Some voice samples below
-
ElevenLabs Launches Free Scribe Speech-to-Text
By
–
ElevenLabs introduced Scribe – a new SOTA speech-to-text model. Now, you can use it for free for a limited time on ElevenLabs to transcribe video and audio files up to 100 MB. https://t.co/eA3m37uSgv pic.twitter.com/F0ht6JF2z2
— 🚨 AI News | TestingCatalog (@testingcatalog) 26 février 2025ElevenLabs introduced Scribe – a new SOTA speech-to-text model. Now, you can use it for free for a limited time on ElevenLabs to transcribe video and audio files up to 100 MB.
-
Google Adds Podcast Conversion for Workspace Docs
By
–
Google now prompts Workspace users to convert their docs into a podcast with NotebookLM. I will need to give it a try "Let's just "listen" to this 50-page RFC instead"
-
GPT-4.5 Research Preview for Pro Users
By
–
BREAKING : GPT-4.5 will likely arrive as a research preview to Pro users. Multiple reports sharing an unintentionally shown notification from the Android app that has been added ahead of the release. h/t @DylanNeve10
-
Google to Add Image Multimodal Output in AI Studio
By
–
Google seems to be gearing up to release image multimodal output on AI Studio. Users will be able to select the media resolution of the resulting images.