Add free trial guide on the top right corner of PapersGPT.
Add redirects to the official API websites for all the LLMs in PapersGPT.
Add review and issue submission entrance on the top right corner of PapersGPT.
v0.4.0
Support the latest SOTA online AI models: Gemini 3 Pro Preview, GPT 5.1, and kimi-k2-thinking.
v0.3.9
Support Linux Zotero MCP.
Support Linux papersgpt agent and local llm.
v0.3.8
Support Claude Haiku 4.5, and free MiniMax M2(free until 2025.11.06) in OpenRouter.
Add local free models: deepseek-distill-llama, mistral-7b, llama-3.1-8b and gemma-3n.
Fix a bug that when chatting multiple PDFs, the output is not consistent with the locale in Zotero.
Fix the bug that Zotero PDF fulltext indexing may be influenced by PapersGPT.
Fix the bug that when inputting your API KEY in PapersGPT on Mac, such as OpenAI API KEY, the LLM can not respond, and output the error log that the API KEY is not provided.
v0.3.7
Support Zotero 8.
v0.3.6
Support intelligent suggestion for authors, tags and titiles of the chatting papers.
Support the new models of Claude Sonnet 4.5, gemini-flash-latest, gemini-flash-lite-latest, deepseek v3.2, glm-4.6.
Change the timeout for the default thinking model gemini-2.5-pro from 30s to 3min.
Support online business models on Linux.
Fix a bug that when papersgpt agent and related libraries were not downloaded completely, It is still normal to use remote business AI models.
v0.3.5
Support the models of o4-mini-deep-research, gpt-4o-mini-search-preview, gpt-4o-search-preview in OpenAI.
Support the models of Grok 4 Fast(free), Tongyi DeepResearch 30B A3B, Qwen3 Next 80B A3B Thinking/Instruct, Kimi K2 0905 in OpenRouter.
Support magistral-medium-latest, magistral-small-latest in Mistral.
Update the latest models of Qwen, Kimi and SiliconFlow.
v0.3.4
Support the fastest responding MCP Server connecting to personal Zotero library in PapersGPT.
v0.3.3
1. Support the official API of Qwen, Mistral, Kimi, Z.ai and SiliconFlow, they are accessed for all the Basic, Premium and Business Users.
2. Fix a bug that the Ollama can't be run normally, as the stream format of Ollama changed, not compatible with OpenAI.
3. Fix a bug that when the internet is not well, the pdf parser may not work.
4. Fix a bug that parsing pdf may be hung or stopped.
5. Fix the bug that the customized selection show wrong.
6. Fix the bug that the title of the plugin show wrong.
v0.3.2
Add OpenAI open source model gpt-oss-20b
v0.3.1
1. Improve the number of PDFs when chatting multiple PDFs, Premium users from 10 to 20; Business users from 20 to 50.
2. Update the latest SOTA models in OpenRouter.
3. Default OpenRouter models are open to Basic users.
4. some default local LLMs can be accessed by Premium users.
v0.3.0
1. The PapersGPT agent has built-in embedding and vector database modules, which help to support very long PDFs and a large number of documents in the backend.
2. Support GPT 5, Claude Opus 4.1, gpt-oss(in OpenRouter), Kimi K2(in OpenRouter), GLM 4.5(in OpenRouter).
3. Fix the bug that the built-in embedding server can not be started normally on Windows.
v0.2.5
1. Support Grok 4 in OpenRouter.
2. Update the latest supported models.
3. Fix the bug for which papersgpt can't be activated in some windows client.
4. Fix the bug that Claude could not be used normally.
v0.2.4
1. 5x Faster PDF Reading.
2. Improve the reading effect of long texts, such as dozens or hundreds of pages of PDF.
v0.2.3
1. Improve the quality of chatting pdf and performance with Local LLM on Mac.
2. Update the latest SOTA Local LLMs(Gemma3 1b, 4b, DeepSeek-Distill-Qwen-8B,...) on Mac.
3. Add new light papersgpt agent for Mac.
4. Fix some bugs of papersgpt agent.
v0.2.2
1. Optimize the startup process of papersgpt agent on Windows.
2. Important notes - The related agents of papersgpt may be mistakenly defined as a Trojan or virus by Microsoft Defender or other anti-virus software. If you want to use the Local LLMs on your Windows, please allow the safe papersgpt agents to run on your device in Microsoft Defender or anti-virus software.
3. Support DeepSeek R1 0528 distill Qwen3 running locally on Windows.
4. Optimize the local LLMs running parameters and improve the answer quality of local LLMs.
v0.2.1
1. Each new version provides an additional 30 free credits for users whose free trial expires.
2. Each credit can be used to chat about PDF once.
3. Add Claude Sonnet 4 and Claude Opus 4.
4. Add Gemini 2.5 Flash Preview 0520 and Gemini 2.5 Pro Preview.
5. Update the hottest models of OpenRouter.
6. Fix the bug that the latest models may not be updated.
7. Fix the bug that when opening the plugin with the configure of local llm, it would not be opened any more.
8. Fix the bug that the showing of Customized is error.
v0.2.0
1. Now support Phi 4 mini reasoning, Phi 4, Gemma 3, Mistral 2503, Qwen 3, DeepSeek Distill Qwen and llama 3.2 on Windows.
2. Models of different sizes are built in based on the size of the local machine GPU.
3. Add papersgpt agent for complex tasks on Windows.
4. Compatible with ollama, if using ollama, please set Customized API URL to http://localhost:11434/api/chat in the selection of Customized.
v0.1.3
1. Add the ability of chatting multiple PDFs in Zotero by PapersGPT. Support generating literature review based on multiple PDFs in Zotero.
2. Support gpt 5, gpt 4.1, o1/o3/o4-mini, gemini-2.5-flash, gemini-2.5-pro and Grok 4(used in OpenRouter).
3. Fix the bug that the selected multiple PDFs may not be read occasionally. Support free Qwen3 in local llm and openrouter.
