The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.
Upstage's[2][3][5] launch post "Solar Pro 4: The Agentic Model That Finishes the Job" is dated August 11, 2026 and its launch promotion runs through September 10; OpenRouter[4] lists the model as released Aug 10, 2026 (14:20 UTC, late evening in Seoul the day before).
Upstage's[1][3][5] developer docs: flagship agentic model, 512K context, 128K max output, English/Korean/Japanese input and output, February 2026 training cut-off, reasoning mode, $0.30 input, $0.06 cached input and $1.20 output per 1 million tokens.
Upstage's[1][2][5] follow-on small model, with only 3B active parameters per token, launched 1 October 2026 beside Solar Pro 4.
Router listing: released Aug 10, 2026, 524K context, currently 70% off at $0.09 input and $0.36 output per 1 million tokens, and Artificial Analysis scores of 28.2 on its Intelligence Index, 89.1% GPQA Diamond and 57.3% Terminal-Bench 2.1.[1]
The predecessor's launch post: IFBench 55.78 (a 52% gain over Solar Pro 2) and Arena Hard v2 62.5 at the same serving efficiency, as the baseline Solar Pro 4 is compared against.[1]
Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.