GPT-5.4
PulseAugur coverage of GPT-5.4 — every cluster mentioning GPT-5.4 across labs, papers, and developer communities, ranked by signal.
- subsidiary of .openai 100%
- developed by .openai 100%
- used by DagsHub 90%
- used by .codex 90%
- instance of large-language models 90%
- uses Molecule.one 90%
- used by Microsoft Mdash 90%
- used by Molecule.one 90%
- developed by Microsoft Research 90%
- competes with Harness-1 90%
- used by Maria 90%
- competes with MAI-Cyber-1-Flash 90%
- 2026-06-19 research_milestone OpenAI and Molecule.one's GPT-5.4 system demonstrated near-autonomous improvement of a drug synthesis reaction. source
- 2026-06-17 research_milestone GPT-5.4 assisted in a medicinal chemistry project, improving yields for key chemical reactions. source
- 2026-05-26 research_milestone An evaluation found GPT-5.4 to be the only model that consistently improved code efficiency when prompted. source
19 day(s) with sentiment data
-
New benchmark HeRE boosts Hebrew relation extraction with LLM denoising
Researchers have developed a novel methodology to create a large-scale benchmark for Hebrew Relation Extraction (RE) using LLM-guided denoising of knowledge graph alignments. This approach addresses the lack of standard…
-
AI developers find fewer tools improve LLM agent reliability
An AI developer found that reducing the number of tools available to an LLM agent significantly improved its reliability. Instead of trying to refine prompts or add more instructions, the developer deleted half the tool…
-
Game Decompilation Leaderboard Shuts Down Amid AI 'Slop' Concerns
The creator of the game decompilation tracking website decomp.dev has shut down its public leaderboard due to the misuse of AI. The site's operator, @encounter, cited instances where AI was used to rapidly complete proj…
-
GitHub Copilot to Deprecate Claude Opus-5, GPT-5.6 Terra Models Nov 16
GitHub Copilot will remove three AI models—Claude Opus-5, Claude Opus 4.8, and GPT-5.6 Terra—from its model picker on November 16, 2026. While these models will still be accessible through their respective vendors' APIs…
-
AI agents generate research papers but lack scientific rigor, study finds
A new study using the ResearchArena scaffold reveals that current AI agents, including Claude Code, Codex, and Kimi Code, can generate complete research papers but struggle with the quality and rigor required for top-ti…
-
New AI model simulates social media behavior using psychology and memory
Researchers have developed PsyMemBehav, a novel system designed to simulate social media behavior by integrating psychological profiles and memory retrieval. This system maps user descriptors and memories to actions lik…
-
AI models rapidly advancing in theoretical science beyond math, study finds
A new study indicates that advanced AI models are rapidly progressing in solving complex theoretical problems across various scientific disciplines, moving beyond mathematics. Researchers found that leading AI models, i…
-
AI graders' accuracy hinges on verified answer keys, studies show
A review of 77 studies on AI grading systems reveals that their reliability is heavily dependent on the quality of the answer key provided. Without a verified reference answer, AI graders tend to err by marking incorrec…
-
Research analyzes ChatGPT citation selection across model versions
A new research paper explores how generative AI models like ChatGPT select which web pages to cite in their responses. The study analyzed four different ChatGPT Web settings, observing how the models rewrite queries, re…
-
Gemini 3.7 Flash leads LLM benchmark for predicting data loss and recovery
A benchmark challenge on Kaggle tested ten large language models on their ability to predict data loss and recovery after destructive file system and database commands. Gemini 3.7 Flash achieved perfect accuracy, correc…
-
New benchmark reveals MLLMs struggle with visual-to-code reproduction
A new benchmark called FigCodeBench has been developed to evaluate the capabilities of Multimodal Large Language Models (MLLMs) in reproducing complex visual figures and generating corresponding code. This framework add…
-
New WOVEN benchmark targets visual reasoning gaps in multimodal LLMs
Researchers have introduced WOVEN, a new benchmark and training dataset designed to improve visual transition reasoning in multimodal large language models (MLLMs). Despite advancements, current MLLMs exhibit significan…
-
New framework uses communication topology to diagnose multi-agent LLM failures
Researchers have developed MAScope, a novel framework designed to diagnose failures in multi-agent Large Language Model (LLM) systems. This system leverages the communication topology between agents to distinguish betwe…
-
New QRAKEN pipeline enhances natural language querying of knowledge graphs
Researchers have developed QRAKEN, a novel neurosymbolic pipeline designed to improve natural language querying of RDF knowledge graphs. Unlike previous methods that rely heavily on schema expectations, QRAKEN grounds i…
-
New Python tool detects LLM agent approval hijack exploits
A new Python checker tool has been developed to identify security vulnerabilities in chained LLM agent skills, specifically focusing on "approval hijacks." This technique exploits how two seemingly harmless skills can c…
-
Study: Cheaper AI models often cost more to run than premium alternatives
A study led by Stanford University revealed that a significant portion of AI models marketed as cost-effective are, in reality, more expensive to operate than their pricier counterparts. The research indicated that Gemi…
-
MiniMax M2.7: Open-source model matches GPT-5.3-Codex on SWE-Pro, rewrites own code
MiniMax has released M2.7, an open-source model with 229 billion parameters, available under the Apache 2.0 license. This model demonstrates significant capabilities, achieving a score of 56.22% on the SWE-Pro benchmark…
-
LLM safety evaluations flawed by distribution shift, research finds
A new research paper highlights significant flaws in how Large Language Model (LLM) safety routing evaluations are conducted. The study, "False Floors: LLM Safety Routing Evaluations Break Under Distribution Shift," dem…
-
GPT-5.4 agent laziness blamed on workflow bugs, not model
An agent built using n8n and GPT-5.4 was observed to perform poorly in production, exhibiting characteristics of "model laziness" such as shorter outputs and less reasoning. However, the issue was not with the model its…
-
LLMs show "Authority Bias", accepting wrong answers from verified sources
A new study reveals that large language models exhibit an