Change log

Feb 2025

• 25 Feb: Playground supports Claude 3.7 Sonnet
• 10 Feb: Playground supports O1 and O3 Mini on Azure
• 07 Feb: Playground supports Gemini 2.0 Flash Lite Preview, Gemini 2.0 Flash, and Gemini 2.0 Pro Exp
• 06 Feb: Models page allows downloading a CSV of all models and their features
• 04 Feb: Playground supports O1 Mini on Azure. gpt-4o-realtime-preview and gpt-4o-audio-preview are available on Azure
• 04 Feb: Allow deploying LLM Foundry even if some model API keys are missing
• 02 Feb: Playground supports Qwen 2.5 Max and Qwen 2.5 Turbo
• 01 Feb: Playground supports O3 Mini

Jan 2025

• 28 Jan: Templates page and Apps usage are paginated and show usage stats
• 26 Jan: Code page has an interface to create temporary tokens
• 26 Jan: Token API supports app to track usage and quota
• 24 Jan: Token API supports expires_in and expires_at
• 22 Jan: Playground supports DeepSeek R1 (reasoning model)
• 22 Jan: Playground supports Gemini 2.0 Flash Thinking Exp 01-21 with code execution
• 22 Jan: LLM Foundry load balances between multiple API keys from the same provider
• 08 Jan: Voyage AI embeddings added

Dec 2024

• 22 Dec: Playground includes Llama 3.3 from Cerebras
• 20 Dec: Playground supports the Gemini 2.0 Flash Thinking model
• 18 Dec: Voicebot tool lets you have audio conversations with LLMs
• 17 Dec: Playground supports RAG for long documents
• 17 Dec: Playground upgraded the Grok model to Grok 2 1212
• 13 Dec: Playground supports a visual display of logprobs (top 5 alternative token choices)
• 11 Dec: Playground supports Gemini 2.0 Flash
• 10 Dec: Playground limits access to safe, private models for specific employees
• 09 Dec: Playground supports private mode
• 07 Dec: Playground supports Llama 3.3 and Gemini Exp 1206
• 05 Dec: Adoption page shows LLM Foundry adoption for the latest month by default
• 04 Dec: Playground supports Amazon Nova
• 02 Dec: xAI models Grok and Grok Vision added via OpenRouter
• 01 Dec: Model picker added to help you select the right model based on features

Nov 2024

• 26 Nov: Playground supports code execution for Gemini models
• 24 Nov: Playground uses the Gemini OpenAI API by default
• 22 Nov: Playground supports 2 new SOTA morals: GPT-4o 2024-11-20 and Gemini Exp 1121
• 22 Nov: Gemini models support OpenAI compatibility
• 21 Nov: Speech API added for dynamic speech generation
• 19 Nov: Playground supports O1-mini and O1-preview
• 19 Nov: Playground supports Qwen 2.5 Coder 32b - a good code model
• 18 Nov: Playground supports Llama 3.2 90b via Google Vertex AI
• 16 Nov: Playground supports Google Drive attachments
• 15 Nov: Playground supports pasting images
• 15 Nov: Adoption page shows LLM Foundry adoption across users
• 12 Nov: Azure Form Recognizer added
• 11 Nov: Log flow shows the flow of recent requests
• 10 Nov: Playground supports Amazon Bedrock models
• 06 Nov: Playground supports grounding via Google Search for Gemini models
• 05 Nov: Playground supports Claude 3.5 Haiku
• 05 Nov: Cerebras models added
• 03 Nov: CORS responses let you access HTTP headers in JavaScript
• 02 Nov: Playground supports Claude 3.5 Sonnet v2 with PDF support

Oct 2024

• 31 Oct: Playground supports document and video attachments for Gemini: PDF, XML, etc and MP4, 3GP, etc.
• 29 Oct: Speak app lets you convert text to speech
• 24 Oct: Playground supports audio input
• 23 Oct: Playground supports Claude 3.5 Sonnet v2. Apps uses Claude 3.5 Sonnet v2 as default
• 18 Oct: Playground supports Llama 3.2 vision models on Groq
• 08 Oct: Playground supports Gemini 1.5 Flash 8b
• 08 Oct: Apps gallery added

Sep 2024

• 28 Sep: PDF API added to convert PDF files to Markdown
• 27 Sep: Playground supports 3 new high quality vision models: Pixtral 12b, Qwen2-VL 72b and Meta Llama 3.2 11b Vision Instruct
• 26 Sep: Playground supports Llama 3.2
• 23 Sep: Proxy API and Markdown API added
• 13 Sep: Playground supports OpenAI's o1-preview and o1-mini
• 07 Sep: Error rate limits added. Any user making more than 5 errors per minute is rate-limited for a minute.
• 07 Sep: OpenRouter models are now supported
• 06 Sep: Playground supports a "# items" option to fetch more articles per page
• 03 Sep: Playground supports drawing diagrams
• 03 Sep: Apps lets you upload images as designs
• 01 Sep: Playground features a "Stop" button to stop generation mid-way

Aug 2024

• 29 Aug: Apps can be saved and support multiple features (e.g. API calls).
• 28 Aug: Gemini 1.5 Experimental models added to Playground
• 27 Aug: Playground supports download as CSV
• 24 Aug: Classify creates networks of documents and topics, simplifying topic modeling
• 21 Aug: Transcribe supports Distil Whisper which transcribes an hour of audio in 15 seconds.
• 21 Aug: chatgpt-4o-latest model added to Playground
• 16 Aug: Playground auto-generates a JSON schema for you
• 15 Aug: Playground lets you specify a JSON schema
• 14 Aug: Playground search works reliably
• 09 Aug: Playground shows code to generate request
• 08 Aug: Playground supports download as Markdown
• 08 Aug: Apps tool added
• 05 Aug: Azure OpenAI model gpt-4o-mini added
• 04 Aug: Deepseek models deepseek-chat and deepseek-coder added

Jul 2024

• 31 Jul: Classify shows a document similarity network
• 26 Jul: Playground supports XML, JSON, and other text file uploads
• 26 Jul: Help pages added
• 24 Jul: Playground supports Llama 3.1
• 22 Jul: Classify auto-discovers topics (via clustering) and names them
• 19 Jul: Playground supports GPT 4o-mini
• 17 Jul: Playground can search the internet or read web pages
• 13 Jul: External users can be added to LLM Foundry
• 08 Jul: Transcribe tool added
• 08 Jul: Playground supports Gemma 2 9b - a frontier model better than Claude 3 Haiku
• 02 Jul: Migrate request / response logs from SQLite to file system
• 02 Jul: Templates page lists all templates
• 02 Jul: Playground can extract text from PDF / DOCX files into user message

Jun 2024

• 24 Jun: Playground supports Claude 3.5 Sonnet on Google Vertex AI
• 23 Jun: Playground supports Claude 3.5 Sonnet
• 05 Jun: Playground disables expensive models after usage of $1.00 / day

May 2024

• 25 May: Token API lets you use LLM Foundry in single-page serverless apps
• 25 May: Azure AI added with Phi-3, Llama-3 models. Also in Playground
• 24 May: Playground Azure has GPT-4o instead of GPT-4
• 21 May: Playground supports Claude 3 Haiku on Google Vertex AI
• 20 May: Playground supports tool usage for Anthropic
• 19 May: Playground supports audio/video files for Gemini
• 18 May: Draw tool added
• 17 May: Google Sheets =LLM() function added
• 15 May: Playground supports Gemini 1.5 Flash
• 14 May: Playground supports images
• 14 May: Playground has GPT-4o instead of GPT-4
• 13 May: Groq models added
• 13 May: Cluster API added
• 12 May: Rewrite tool added
• 11 May: Extract tool added
• 06 May: Usage shows usage by app as well as email
• 03 May: Playground FAQ explains what LLM Foundry is

Apr 2024

• 29 Apr: Classify tool examples added
• 28 Apr: Registered apps added
• 27 Apr: Classify tool added
• 26 Apr: Similarity API supports image embeddings via multimodalembeddings
• 25 Apr: Usage is updated every 2 hours, shows usage by email and by app, and cost
• 21 Apr: Microsoft auth supported for Straive users too
• 21 Apr: Google Vertex AI added
• 20 Apr: Similarity API added
• 20 Apr: Playground now has the llama-3-8b-instruct (instead of starling-lm-7b-beta)
• 12 Apr: Gemini added
• 11 Apr: Template API added
• 11 Apr: Azure uses the Straive Azure tenant instead of Gramener's
• 09 Apr: Token API added
• 04 Apr: Cloudflare Workers AI added -- notably SQLCoder
• 01 Apr: Usage supports filter by date

Mar 2024

• 29 Mar: LLMFoundry launched, rebranded from LLM Proxy
• 29 Mar: Anthropic added
• 28 Mar: Usage stats added
• 27 Mar: Playground supports templates
• 25 Mar: Playground supports 🔑variables
• 22 Mar: Playground renders Markdown while streaming
• 19 Mar: Playground supports streaming
• 19 Mar: Playground added
• 11 Mar: Azure supports gpt-4-vision-preview
• 07 Mar: History added

Feb 2024

• 25 Feb: Log out added
• 24 Feb: Usage shows users and apps
• 17 Feb: Usage shows visualizations
• 04 Feb: Azure OpenAI added
• 03 Feb: OpenAI API allows use of OpenAI keys as well as LLMFoundry