GPT-6 Sol
GPT‑6 Sol, also called GPT‑6.1 Sol, is a cheaper Astra‑like model released earlier in 2026. It uses the same base weights as the flagship GPT‑6 Sol but performs two inference passes instead of three, enabling quicker output while keeping similar accuracy. The design details were disclosed in a Microsoft technical write‑up that leaked the looped‑inference mechanism—loops the same transformer layers multiple times before producing output. ScaledownSummarizationMiddleware USES_MODEL openai:gpt-6-sol Gemini 4 Argon model has the same price as GPT-6 Sol OpenAI GPT‑5.6 models are available through the Amazon Bedrock Converse API. OpenAI claimed that GPT-6 Sol scored about 21 percent accuracy at low reasoning effort OpenAI has the GPT-6 Sol model· Entity summary generated by Graphiti
Relation timeline
BY VALID-FROM DATE- OpenAIHAS MODELGPT-6 Sol
OpenAI has the GPT-6 Sol model
- OpenAICLAIMEDGPT-6 Sol
OpenAI claimed that GPT-6 Sol scored about 21 percent accuracy at low reasoning effort
- Gemini 4 ArgonSAME PRICE ASGPT-6 Sol
Gemini 4 Argon model has the same price as GPT-6 Sol
- ScaledownSummarizationMiddlewareUSES MODELGPT-6 Sol
ScaledownSummarizationMiddleware USES_MODEL openai:gpt-6-sol
- GPT‑5.6AVAILABLE THROUGHGPT-6 Sol
OpenAI GPT‑5.6 models are available through the Amazon Bedrock Converse API.
Relations are facts Graphiti extracted from reports, not verbatim quotes. An end date means later reporting showed the relation no longer holds.
Events 4
- Can AI help solve the COBOL code crisis?1 cited report · 2026-10-26
- Access OpenAI GPT models on Amazon Bedrock from Canada with cross-Region inference1 cited report · 2026-10-21
- llm-gemini 0.351 cited report · 2026-10-14
- Building Long-Running Agents with ScaleDown and Deep Agents1 cited report · 2026-10-13
Source reports 4
- Can AI help solve the COBOL code crisis?The Verge · AI · 2026-10-26
- Access OpenAI GPT models on Amazon Bedrock from Canada with cross-Region inferenceAWS Machine Learning Blog · first-party · 2026-10-21
- llm-gemini 0.35Simon Willison · 2026-10-14
- Building Long-Running Agents with ScaleDown and Deep AgentsLangChain Blog · 2026-10-13