o1-mini
2024-09-12 · OpenAI A smaller reasoning model released alongside o1-preview with a focus on mathematics and coding. It had its own identifier and was not an access tier of the preview model. Source
MODEL DIRECTORY
New releases, in one place. Explore models by their publication date.
2024-09-12 · OpenAI A smaller reasoning model released alongside o1-preview with a focus on mathematics and coding. It had its own identifier and was not an access tier of the preview model. Source
2024-09-12 · OpenAI A preview model for multi-step reasoning in mathematics, science and coding. It is recorded separately from the full o1 release in December. Source
2024-09-05 · DeepSeek This release combined general conversation and coding in one model, reducing the need to switch between specialists. This is the September version, not the December update. The…
2024-07-18 · OpenAI A small multimodal model with text and image inputs, suited to high-volume classification, extraction and short answers. It was distinct from GPT-4o. Source
2024-06-28 · Baidu This ERNIE 4.0 upgrade was announced at WAVE SUMMIT on June 28. General availability for enterprises followed on July 5. The model name stands in for an unverified API identifier.…
2024-06-05 · Zhipu AI The 9B-scale open-weight GLM-4 family supported local deployment and adaptation. This date marks the weights release, not the earlier commercial GLM-4 launch. The model name is…
2024-05-15 · ByteDance A general model in the Doubao family presented by Volcano Engine for enterprise questions and text generation. A specific API snapshot or context tier has not been verified. The…
2024-05-14 · Google This lightweight multimodal model entered public preview for fast summarization, classification and extraction. It was distinct from Gemma 2, teased at the same event. The model…
2024-05-13 · OpenAI This multimodal model worked across text, vision and audio. Features rolled out in stages; a live voice demo did not mean all API capabilities were available on day one. Source
2024-05-06 · DeepSeek A general mixture-of-experts model that activated only some experts per token to limit inference cost. The base and chat variants served different purposes. The model name is…
2024-04-17 · MiniMax MiniMax announced this family of general language models for chat, writing and text understanding. The date applies to the family, not one endpoint configuration. The model name…
2024-03-13 · Anthropic This lightweight variant targeted low-latency chat, classification and information extraction. It was announced with the family on March 4; the date here marks availability.…
2024-03-04 · Anthropic Sonnet occupied the middle tier of Claude 3, balancing speed with performance on complex tasks. It became available alongside Opus; Haiku followed later. Source
2024-03-04 · Anthropic This Claude 3 variant handled complex reasoning, analysis and writing with text and image inputs. It became available on the day of the announcement. Source
2024-02-15 · Google This long-context multimodal model could answer questions about documents, code, audio and video. Its limited preview did not grant every developer the same quota. The model name…
2024-01-16 · Zhipu AI Zhipu announced this general model at its technology open day for question answering, coding and tool use. It is distinct from the 9B weights released in June. Source
2023-12-06 · Google The multimodal family comprised Ultra, Pro and Nano. At announcement, Pro powered Bard and Nano supported some Pixel 8 Pro features; Ultra was not publicly available that day. The…
2023-11-29 · DeepSeek A general Chinese-English family with 7B and 67B sizes, each available in base and chat variants. It is distinct from the code-oriented DeepSeek Coder. The model name is used…
2023-11-21 · Anthropic Its 200,000-token context window supported analysis of long reports or code. Input capacity did not guarantee accurate recall of every detail. Source
2023-11-06 · OpenAI This DevDay API preview supported longer contexts for documents and multi-turn conversations. It was not the later stable Turbo release. Source