GPT-5.6 Sol (AWS Mantle)

According to a third-party technical guide published on August 26, 2026, OpenAI deployed its next-generation flagship model GPT-5.6 Sol in late August 2026 with a new "Ultrafast Inference Mode" that targets a time-to-first-token (TTFT) below 180ms while preserving o1-class multi-step planning and mathematical reasoning The same guide reports that alongside the model deployment OpenAI began testing a ChatGPT Ads monetization layer across the UK, Brazil, Mexico, Japan, and South Korea, with contextually relevant sponsor cards shown to free-tier users while Pro and Plus subscribers remain ad-free. The guide also presents a benchmark mat

OpenAI announced on August 11, 2026 that its Daybreak models are now available on AWS through Amazon Bedrock, with GPT-5.6 Sol explicitly named as one of the frontier general-purpose models accessible under the Daybreak Blue access level. Daybreak Blue is described as providing "frontier general-purpose models, includi From a developer-integration standpoint, eligible customers access GPT-5.6 Sol on AWS via the Amazon Bedrock console or through the Responses API using the bedrock-mantle endpoint, directly aligning with the AWS Mantle routing layer that LLM Gateway exposes for this model. The announcement positions Bedrock as the ente

The BenchmarkList leaderboard page for OpenAI GPT-5.6 Sol reports concrete agentic-evaluation scores with explicit dates: BrowseComp 92.2% (X-high effort, launch post dated 2026-07-09) and 90.4% at Max reported effort (2026-07-16 source, 2026-08-28 entry), making GPT-5.6 Sol the current field leader on BrowseComp with BenchmarkList lists GPT-5.6 Sol at OpenAI's standard list price of $5.00 per million input tokens and $30.00 per million output tokens, identical to the pricing repeated in other third-party sources. Across the 39 agentic evaluations summarized, the model sits at the 88th percentile median and has scored a 1st-place ra

OpenAI unveiled Ultrafast on August 13, 2026, a new preview service tier that runs GPT-5.6 Sol on Cerebras wafer-scale silicon at up to 750 output tokens per second, which OpenAI describes as up to 14 times faster than its Standard processing tier. The tier is launching first in the OpenAI API and is currently invite-o OpenAI positions Ultrafast for latency-sensitive workloads including incident response, financial research and security analysis, real-time customer support and voice, commerce personalization, and interactive live research where overnight batch runs are no longer required. The speed and latency figures cited are OpenA