News
ResearchOpenAI pens AI safety agreement with four firmsReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierOpenAI鈥檚 New Hacking Model Found Dozens Of Holes In Apple TechFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIResearchOpenAI pens AI safety agreement with four firmsReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierOpenAI鈥檚 New Hacking Model Found Dozens Of Holes In Apple TechFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAI
EveryLLM

Research access only

Shared with researchers on request, or described in a paper only. 58 models on EveryLLM so far.

Some models were never offered to the public at all. They were described in a paper, shown in a demonstration, or shared with selected researchers, testers or partners. Labs keep a model back when it is an experiment, when it is too costly to serve, or when they judge its risks too high for wide release. Research-only models still shape the field through their papers and the techniques that others copy.

Papers without products

DeepMind's Gopher (December 2021) and Chinchilla (March 2022) were known almost entirely through their papers. Google's LaMDA was shown at Google I/O in May 2021 and offered to a small group through the AI Test Kitchen app in 2022. DeepMind's Sparrow (2022), a dialogue agent trained with human feedback and a set of rules, and Google's VideoPoet (December 2023) were likewise research systems.

Held back for safety

Labs sometimes cite risk. OpenAI staged the release of GPT-2 in 2019 over misuse fears. In April 2026 Anthropic gave its Claude Mythos Preview model only to selected companies, under the name Project Glasswing, to scan critical software for vulnerabilities, citing the model's skill at finding them. Critics of such limits argue that they concentrate power in a few hands; supporters say staged access lets defenders prepare first.

Famous research only models

Further reading

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.