Today's price · per 1M tokens
Input
$0.300
per 1M tokens
Output
$1.20
per 1M tokens
Blended
$0.525
blended $/1M — 3:1 weighted input:output
Cached input
$0.060
20% of input — prompt caching
How this price is scoped: List price tracked, per 1M tokens, exclusive of 10% VAT: input $0.30, cached input $0.06, output $1.20, as Upstage's model page prints it. Upstage's pricing page (upstage.ai/pricing/api) has extended the launch discount rather than ending it on Oct 10: 70% off ($0.09 / $0.018 / $0.36) through Oct 19 2026 (UTC), then 50% off ($0.15 / $0.03 / $0.60) from Oct 20 2026 with no end date given. The price you pay is the discounted one until Upstage ends the offer.
Price receipt
Partial receipt. We read Upstage's own pricing page on , and every price above is printed on it — for each one, Solar Pro 4 is the nearest model name on the page and no other model we track sits closer. What that capture does not give us is an entry for this model we can parse, so the match rests on where the numbers sit on the page rather than on a row we could read end to end. We keep the page text we read; its content hash is a51ade37ae. Check the provider's page before relying on it.
Overview
Upstage's agentic Solar flagship, successor to Solar Pro 3, with a 512K context window and up to 128K output tokens. Upstage's list rate is $0.30/$1.20 per 1M tokens, and its pricing page currently sells it at a discount from that list price.
Capabilities
struck through = not supportedBenchmark performance
accuracy % · higher is betterSources: Upstage Solar Pro 4 launch blog (GPQA Diamond) · Upstage Solar Pro 4 launch blog (SWE-Bench Verified, OpenHands in-house harness) (SWE-Bench Verified) · Artificial Analysis (Terminal-Bench 2.1) · Upstage Solar Pro 4 launch blog (LiveCodeBench) (LiveCodeBench) · Upstage Solar Pro 4 launch blog (MCP-Atlas) (MCP Atlas) · Upstage Solar Pro 4 launch blog (BrowseComp, in-house harness) (BrowseComp)
- AA Intelligence Index: 14.2 (Artificial Analysis composite across reasoning, knowledge and coding evals)
- AA-Omniscience Index: 1.4 (−100–100)
- GDPval-AA v2: 1105.7 ELO (Real-world work tasks, human baseline = 1000)
- Arena (arena.ai): 1387.5 ELO (Human preference)
- Arena WebDev: 1369.3 ELO (Human preference on building web apps)
- Arena Agent: -0.226 (Score from real agent sessions (IPS); higher is better)
- Arena Agent cost per task: $0.04 (Median measured cost of one real Agent Arena task)
- Humanity's Last Exam: 29.8%
- GPQA Diamond: 89.1%
- AA-LCR (long-context reasoning): 70.7%
- CritPt (research-level physics): 6.3%
- Terminal-Bench 2.1: 57.3%
- Terminal-Bench 4.0: 0%
- SciCode: 48%
- Cost to run the AA Intelligence Index: $0.956 per task, 63.5 min per task
- Measured by AA: 81.8 tokens/sec output, 2.27s to first token, 35.22s to first answer token
- Ranks #73 of 100 benchmarked models by average score
- Ranks #119 of 189 comparably-measured models by percentile score, across 5 independent measurements
- Ranks #38 of 189 comparably-tested models by normalized performance per dollar
- Strongest at Terminal-Bench 2.1 — 57%, #36 of 46
81.8 tokens/sec output 2.27s latency to first token (TTFT)
Throughput: Artificial Analysis, measured 2026-11-14 · Source: Artificial Analysis (artificialanalysis.ai)
Source: Artificial Analysis, LMSYS Chatbot Arena (UC Berkeley), Upstage Solar Pro 4 launch blog · updated · See full rankings →
Specifications
- Provider
- Upstage
- Context window
- 524K tokens
- Modality
- text
- Parameters
- Proprietary
- Open source
- No — proprietary
- Released
- Status
- Current
- Last updated
- Tags
Availability verified: — listed on Upstage's own page