GPT-5.5
GPT-5.5 is a multimodal large language model released by OpenAI, belonging to the GPT series, with support for both text and image inputs.
Core Facts
Timeline (last 90 days)
The multi-model workspace adopted by SmophyAI can send a single question simultaneously to GPT-5.5, Claude Opus 4.8, Gemini, Grok, Perplexity and DeepSeek, display the results side by side in the same window, and each model is connected to real-time web data through Brave Search.
Claude's output in long-form writing (such as a 3000-word article) is more natural than GPT-5.5's, and the gap accumulates as the length increases.
In William Liu's experiment, across all 31 test samples, the token counts reported by the seven GPT models were completely identical, all at 44,794.
William Liu compared the tokenization results of multiple GPT models using a set of fixtures, with the experiment covering seven GPT models, including GPT-5.5, three variants of GPT-5.6 (Sol/Terra/Luna), and three variants of GPT-6 (Astra/Sol/Luna).
Verdandi uses the GPT-6-luna model, and ACU uses the GPT-5.5 and text-embedding-3 models.
GPT-5.5 ran out of usage quota, leaving 35 of the 90 questions unfinished; all 55 answered questions were correct, but Kaggle counted the unanswered questions as wrong, causing its official score to be lower.
GPT-5.5 led the benchmark with an overall score of 0.94, a director's note score of 0.94, and a story quality score of 0.85.
GPT-5.5 rewrote every shot when shifting time periods, and only modified the necessary shots when cutting the budget or merging scenes, demonstrating appropriate behavior.
In the first run, Claude Opus 5 scored only 0.08 and GPT-5.5 errored out entirely; upon investigation this was found to be a 403 PermissionDeniedError quota-exceeded error, where concurrent requests burned through the quota for expensive models and the scorer incorrectly treated errored calls as wrong answers.
Jev's operating cost is approximately 65 times lower than frontier models such as GPT-5.5
22 more timeline events
All Facts (20)
GPT-5.5是OpenAI于2025年中期发布的模型，相比GPT-4o在推理深度和上下文理解方面有显著提升
90%VerifiedGPT-5.5 was released at the end of April 2025.
80%VerifiedGPT-5.5据称是目前所有大模型中综合实力最强的
80%VerifiedAI models capable of vulnerability discovery may allow people without deep security knowledge to find and exploit vulnerabilities, lowering the barrier to attacks.
75%UnverifiedGPT-5.5 xhigh 是 OpenAI 模型系列中的高推理能力配置，'xhigh'表示模型在生成回答时会投入更多计算资源进行深度推理
85%UnverifiedGPT-5.5属于OpenAI，Claude属于Anthropic，Gemini属于Google，三者原生API格式各有差异
85%UnverifiedGPT-5.5 xhigh中的xhigh指的是计算预算等级，意味着模型在生成代码时会投入更多的推理计算资源
80%UnverifiedThe public availability of AI models with professional-grade vulnerability discovery capabilities may lower the technical barrier to cyberattacks.
70%Unverified基础大模型能力的提升正在直接转化为垂直领域应用的质量跃升
70%UnverifiedLlama 3.1 405B's score on MMLU is already close to the level of GPT-4o, and Qwen 2.5 72B has surpassed GPT-4 Turbo on certain metrics in math and code generation benchmarks.
60%Unverified新一代结合GPT-5.5的Codex能够理解跨文件的依赖关系、识别异步竞态条件等复杂bug，以及评估算法复杂度
60%UnverifiedClaude's output in long-form writing (such as a 3000-word article) is more natural than GPT-5.5's, and the gap accumulates as the length increases.
50%UnverifiedIn William Liu's experiment, across all 31 test samples, the token counts reported by the seven GPT models were completely identical, all at 44,794.
50%UnverifiedVerdandi uses the GPT-6-luna model, and ACU uses the GPT-5.5 and text-embedding-3 models.
50%UnverifiedGPT-5.5 ran out of usage quota, leaving 35 of the 90 questions unfinished; all 55 answered questions were correct, but Kaggle counted the unanswered questions as wrong, causing its official score to be lower.
50%UnverifiedGPT-5.5 rewrote every shot when shifting time periods, and only modified the necessary shots when cutting the budget or merging scenes, demonstrating appropriate behavior.
50%UnverifiedGPT-5.5 led the benchmark with an overall score of 0.94, a director's note score of 0.94, and a story quality score of 0.85.
50%UnverifiedIn a test of 1,000 messages from the banking77 dataset, Jev achieved an accuracy of 80%, Gemini Flash 78.5%, and GPT-5.5 84.9%
50%UnverifiedUsing a model cascade pattern, escalating messages where Jev's confidence is below 0.7 to GPT-5.5 raised accuracy to 83.7%, with cost increasing from $0.07 to approximately $2.58, which is lower than the 84.9% accuracy and $4.65 cost of using GPT-5.5 throughout
50%UnverifiedJev's operating cost is approximately 65 times lower than frontier models such as GPT-5.5
50%