LLMcompare

Search

Search for a command to run...

A · open · Solar

Solar Pro 3

Upstage

B · closed · Solar

Solar Pro 4

Upstage

0:9

overall capability wins (0 ties across 9 scored). Solar Pro 4 leads overall.

  • Reasoning0:4
  • Coding0:2
  • Tool use0:1
  • Pricing3:0
  • Specs0:2

Verdict

Solar Pro 4 leads coding (widest gap: +45.3 on TermBench); Solar Pro 4 offers ~4.0x larger context (512K vs 128K); Solar Pro 4 allows ~15.6x more max output (128K vs 8K).

  • Solar Pro 4 leads coding (widest gap: +45.3 on TermBench)
  • Solar Pro 4 offers ~4.0x larger context (512K vs 128K)
  • Solar Pro 4 allows ~15.6x more max output (128K vs 8K)
  • Solar Pro 3 ships open weights (Apache 2.0)
  • Solar Pro 3 is ~2.0x cheaper on a blended token basis than Solar Pro 4

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

Solar Pro 3
$0.30
Solar Pro 4
$0.60

Solar Pro 3 is about 2.0x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

Upstage

Upstage

Family

Solar

Solar

License

Apache 2.0

Proprietary

Open weights

Yes

No

Release

Jan 26, 2026

Aug 10, 2026

Knowledge cutoff

-

-

API / provider

Upstage

Upstage

Modalities

text → text

text → text

Specs

Context window

128K

512K

B +384K

Max output

8K

128K

B +120K

Parameters

102B (12B act.)

—

—

Pricing

Input $/1M

$0.15

$0.30

A +$0.15

Output $/1M

$0.60

$1.20

A +$0.60

Blended $/1M (3∶1)

$0.26

$0.52

A +$0.26

Speed

tok/s

90

-

—

TTFT (s)

0.3

-

—

Reasoning

MMLU-Pro

—

—

—

GPQA Diamond

72.4

89.1

B +16.7 pts

Humanity's Last Exam

10.3

29.2

B +18.9 pts

AIME 2025

—

—

—

MATH-500

—

—

—

Humanity's Last Exam (with tools)

—

—

—

AA-Omniscience Accuracy

18.5

18.9

B +0.4 pts

AA-LCR v1.1

32.3

74.0

B +41.7 pts

CritPt

—

5.4

—

MMMU-Pro

—

—

—

IFBench

71.2

—

—

Chartography

—

—

—

Chartography (With Tools)

—

—

—

Coding

SWE-bench Verified

—

—

—

SWE-bench Pro

—

—

—

SWE-bench Multilingual

—

—

—

LiveCodeBench

—

—

—

Terminal-Bench 2.1

12.0

57.3

B +45.3 pts

Aider Polyglot

—

—

—

Terminal-Bench 3

—

—

—

BigCodeBench

—

—

—

SciCode

25.5

44.6

B +19.1 pts

CursorBench

—

—

—

SWE-Rebench

—

—

—

NL2Repo-Bench

—

—

—

DeepSWE

—

—

—

WebDev Arena

—

1,372

—

Terminal-Bench 4.0

—

—

—

CursorBench 4.0

—

—

—

FrontierCode v1.1 (Main)

—

—

—

Arena

LMArena Elo

—

1,376

—

Solar Pro 3

Context
128K / 8K out
Parameters
102B (12B act.)
Price
$0.15 / $0.60
Speed
90 tok/s · 0.3s TTFT
Modalities
text → text
License
Apache 2.0

Upstage's 102B MoE (12B active) — strong Korean/English bilingual model with competitive sub-$1 blended API pricing.

Solar Pro 4

Context
512K / 128K out
Parameters
—
Price
$0.30 / $1.20
Speed
—
Modalities
text → text
License
Proprietary

Upstage's proprietary successor to Solar Pro 3 — 512K context with up to 128K output tokens for loading multiple long documents into one session; EN/KO/JA text, released Aug 2026.

Benchmark charts

Winner bars are emphasized. Per-benchmark deltas sit above each chart.

Reasoning & knowledge

GPQA: B +16.7HLE: B +18.9AA-LCR: B +41.7Omniscience Acc.: B +0.4

Coding

TermBench: B +45.3SciCode: B +19.1

Tool use & function calling

τ³-Banking: B +14.6

Pick a different pair · Back to catalog