UpstageReleased Jul 22, 2026
Open weights

Solar Open2 250B

Its best results are in Academic Reasoning (#92 of 223) and Reliability (#90 of 188). Five of seven dimensions are estimated, because no results are published yet.

Composite IQ ?
102
#126 of 186 ranked models · ahead of 32%
World (Other)'s #11 model, behind mistral-large-4 (118)
Effective cost
Not priced
No standard API rate
Speed
Not measured
No speed data yet
Context
1Mtokens
1,048,576 tokens in
Evidence
2of 7 dims
6 of 39 benchmarks

IQ profile

Seven dimensions, averaged equally into Composite IQ
Compare shape with
100 avg human130160● Abstract88#130 · est.● Math97#141 · est.● Academic123#92● Coding92#172 · est.● Computer Use104#113 · est.● Reliability113#90● Knowledge Work97#137 · est.
MeasuredEstimated95% range
#1 in dimensionTop-10 median

IQ dimensions

Ranks are among models scored in each dimension. Ranges are model-based estimates, not calibrated confidence intervals.

DimensionIQPosition · 70–160RankEvidencevs Top 10
88estimated
#130 / 221
0/5
-44.5
97estimated
#141 / 186
0/5
-52.5
123measured
#92 / 223
4/6
-12
92estimated
#172 / 187
0/6
-59
104estimated
#113 / 186
0/5
-42.5
113measured
#90 / 188
2/6
-22.5
97estimated
#137 / 186
0/6
-35
Solar Open2 250B95% range (model-based, uncalibrated)Top-10 median#1 in dimension100 = average human

Benchmark results

6 of 39 Composite IQ benchmarks have a published result for Solar Open2 250B. Benchmarks without a result stay listed so gaps are visible.

Abstract Reasoning IQ 88 · estimatedNo published results0/5
No published results yet for Abstract Reasoning. The 88 shown is estimated from Solar Open2 250B's other dimensions and will be replaced as results come in. How estimates work
ARC-AGI-3ARC-AGI-2ARC-AGI-1Chess PuzzlesMystery Game Puzzles
Mathematical Reasoning IQ 97 · estimatedNo published results0/5
No published results yet for Mathematical Reasoning. The 97 shown is estimated from Solar Open2 250B's other dimensions and will be replaced as results come in. How estimates work
FrontierMath Tier 4FrontierMath Tier 1-3ProofBench 1.1MathArenaAIME
Academic Reasoning IQ 123 · measuredBest: #74 of 167 on SciCode4/6
BenchmarkScorePercentileRankLeader · source
SciCode
Artificial Analysis v4.3.2 · Oct 5
48.0%
#74 / 167
Leader opus-5.5 · 66.9%
Artificial Analysis v4.3.2 · Oct 5
CritPt
Artificial Analysis v4.3.2 · Oct 5
5.7%
#78 / 159
Leader Muse Wonder Full · 32.9%
Artificial Analysis v4.3.2 · Oct 5
GPQA Diamond
Artificial Analysis v4.3.2 · Oct 5
85.7%
#79 / 147
Leader gpt-6-astra · 96.1%
Artificial Analysis v4.3.2 · Oct 5
Humanity's Last Exam
Artificial Analysis v4.3.2 · Oct 5
28.5%
#92 / 167
Leader opus-5.5 · 61.4%
Artificial Analysis v4.3.2 · Oct 5
Not yet reported
MMLU-ProMMMU-Pro
Programmatic Reasoning IQ 92 · estimatedNo published results0/6
No published results yet for Programmatic Reasoning. The 92 shown is estimated from Solar Open2 250B's other dimensions and will be replaced as results come in. How estimates work
LiveCodeBenchIOI V2Terminal-Bench 4.0ProgramBenchSWE-rebenchMirrorCode
Computer Use IQ 104 · estimatedNo published results0/5
No published results yet for Computer Use. The 104 shown is estimated from Solar Open2 250B's other dimensions and will be replaced as results come in. How estimates work
BrowseCompOSWorld-VerifiedMCP AtlasArena.ai Agent ArenaAgents' Last Exam
Reliability IQ 113 · measuredBest: #76 of 158 on AA Omniscience2/6
BenchmarkScorePercentileRankLeader · source
AA Omniscience
Artificial Analysis v4.3.2 · Oct 5
-1.8
#76 / 158
Leader opus-5.5 · 46.4
Artificial Analysis v4.3.2 · Oct 5
AA Long Context Reasoning v1.1
Artificial Analysis v4.3.2 · Oct 5
71.7%
#104 / 154
Leader kimi-k3 · 88.7%
Artificial Analysis v4.3.2 · Oct 5
Not yet reported
SimpleQA VerifiedBullshitBench v2IFBenchFACTS Grounding
Knowledge Work IQ 97 · estimatedNo published results0/6
No published results yet for Knowledge Work. The 97 shown is estimated from Solar Open2 250B's other dimensions and will be replaced as results come in. How estimates work
AA-Briefcase v1.1GDPval-AA v2.1AutomationBench-AAGDP.pdfAA-AnalystAgentEnterpriseOps-Gym-AA

Release & availability

Facts from the publisher's announcement and documentation.

Released
Developer
Upstage
Access
Open weights on Hugging Face for local deployment
Weights
Open weightsDownloadable and self-hostable
Inputs
text
Context
1,048,576 tokens
API price
No per-token rateNo current standard hosted token price has been verified for this checkpoint; downloadable weights are not a zero-dollar hosted API.

What Upstage says

  • Solar Open2 is a 250B-parameter mixture-of-experts model with about 15B active parameters.
  • The model supports roughly 1M tokens of context and English, Korean, and Japanese.

Limitations

  • Weights use the Upstage Solar License; review its conditions for deployment.
  • The publisher reports deployment on two H200 GPUs with quantization. No standard hosted price is verified.

Publisher statements, not AI IQ measurements.

About this score

How much of Solar Open2 250B's IQ is measured, and how the rest is estimated.

6/ 39benchmark results
2/ 7dimensions measured
5estimated
0/5Abstract
0/5Math
4/6Academic
0/6Coding
0/5Computer Use
2/6Reliability
0/6Knowledge Work
Bar height = share of benchmarks with resultsEstimatedNo score
Measured results are shown as published

Every score in the benchmark tables comes from its source unchanged. Composite IQ averages all seven dimensions equally.

Estimates fill the gaps

Abstract Reasoning, Mathematical Reasoning, Programmatic Reasoning, Computer Use and Knowledge Work have no published results yet, so their IQ is estimated from Solar Open2 250B's other dimensions.

A missing result is not a zero

Benchmark versions and evaluation harnesses can differ between sources, so small gaps between models may not be meaningful.

Read the scoring methodology →

Compare Solar Open2 250B

Side-by-side pages for the models it is most often weighed against.

Solar Open2 250BComparison model
Browse all comparisons →

Frequently asked questions

Quick answers about Solar Open2 250B, drawn from the same data as the rest of this page.

How AI IQ scores models →Get this data from the API →
What is the IQ of Solar Open2 250B?
Solar Open2 250B has an estimated IQ of 102, which ranks #126 out of 186 public models on the AI IQ leaderboard. The score combines seven equally weighted reasoning dimensions using measured benchmark results and conservative estimates for missing coverage.
What is Solar Open2 250B best at?
Among the dimensions with published benchmark results, Solar Open2 250B ranks highest in Academic Reasoning (IQ 123, #92 of 223) and Reliability (IQ 113, #90 of 188).
What is the EQ of Solar Open2 250B?
Solar Open2 250B doesn't have an EQ score yet. EQ (emotional reasoning) is calculated from EQ-Bench 4 and AttuneBench, and neither has published a result for Solar Open2 250B. AI IQ doesn't estimate EQ from other benchmarks, so the score appears once one of those results is published.
Which benchmarks has Solar Open2 250B been tested on?
Solar Open2 250B has published results on 6 of the 39 benchmarks behind AI IQ's Composite IQ, including SciCode, AA Omniscience, CritPt, GPQA Diamond and Humanity's Last Exam.
How much does Solar Open2 250B cost?
AI IQ doesn't have a verified standard per-token API price for Solar Open2 250B, so it can't calculate an effective cost. This is common for models offered only through subscriptions, restricted-access programs, or self-hosting.
Is Solar Open2 250B open source?
Yes — Solar Open2 250B is an open-weights model: its trained weights are published, so it can be downloaded and self-hosted.
Who developed Solar Open2 250B?
Solar Open2 250B was developed by Upstage, based in South Korea.
When was Solar Open2 250B released?
Solar Open2 250B was publicly released on July 22, 2026.
How large is the context window of Solar Open2 250B?
Solar Open2 250B supports a context window of 1,048,576 tokens — the amount of text it can consider in a single request.
How fast is Solar Open2 250B?
AI IQ doesn't have independent speed measurements for Solar Open2 250B yet. Speed comes from independent tests of a model's hosted API, which aren't available for every model — especially new, restricted-access, or self-hosted ones.