Solar Pro 4

Enera's August 23, 2026 field note confirms Upstage AI officially launched Solar Pro 4 on August 20, 2026, citing a PR Newswire announcement and independent coverage by The New Stack. It quotes Upstage's Kasey Roh directly: "Save the frontier models for the frontier problems. We built the workhorse," positioning Solar The post reports that within a week of Solar Pro 4 appearing on OpenRouter, its token consumption exceeded 370 billion tokens, signaling rapid developer adoption. It highlights the same production failure modes the model targets, including schema drift in tool calls and inconsistent instruction-following across multi-s

The New Stack (August 20, 2026) reports on Solar Pro 4's official launch, framed by Upstage's Head of US Kasey Roh as a "workhorse" agent reliability model positioned below frontier systems. Roh argues that pointing flagship models at repetitive production work is "overkill and honestly a liability" because teams pay f The article emphasizes behavioral reliability as Solar Pro 4's differentiator: instruction-following that holds across multi-step agent chains, tool-call structure that does not drift out of schema, and agentic execution confined to defined policy boundaries. Upstage designed the model for enterprise workloads such as

According to Artificial Analysis (published August 12, 2026), Korean AI lab Upstage has released Solar Pro 4 as its new proprietary flagship reasoning model, replacing Solar Pro 3 from April 2026. The model scores 42 on the Artificial Analysis Intelligence Index, a 27-point jump over Solar Pro 3's 14, placing it alongs The analysis highlights major agentic and long-context gains: Terminal-Bench v2.1 rises from 12% to 57%, AA-LCR from 31% to 71%, τ³-Banking from 9% to 23%, and GDPval-AA Elo climbs from 498 (well below the 1000 human baseline) to 1277. Solar Pro 4 uses a 384K context window with 256K max output, supports text input/out

A Baidu encyclopedia entry on Solar Pro 4 describes it as a 2026 commercial large model from South Korean AI lab Upstage, officially announced on August 13, 2026, and positioned for complex AI agent tasks. The entry notes that Upstage has not disclosed the parameter count for Solar Pro 4 in its official blog post. The entry reports a 512K-token context window with a 128K-token maximum output length, allowing users to load multiple contracts, reports, and data files in a single session. Its primary improvements over prior Upstage models focus on agent evaluations that mirror real work scenarios, including long-document handling,