LLMcompare

Search

Search for a command to run...

A · open · DeepSeek

DeepSeek V4 Pro

DeepSeek

B · open · GLM

GLM-5.2

Zhipu AI

4:12

overall capability wins (0 ties across 16 scored). GLM-5.2 leads overall.

  • Reasoning2:4
  • Coding0:5
  • Arena0:1
  • Tool use0:2
  • Composite indices0:2
  • Pricing3:0
  • Speed0:2
  • Specs2:0

Verdict

GLM-5.2 leads coding (widest gap: +146.1 on WebDev Elo); GLM-5.2 ranks higher on LMArena (+15 Elo); DeepSeek V4 Pro offers ~5.0x larger context (1M vs 200K).

  • GLM-5.2 leads coding (widest gap: +146.1 on WebDev Elo)
  • GLM-5.2 ranks higher on LMArena (+15 Elo)
  • DeepSeek V4 Pro offers ~5.0x larger context (1M vs 200K)
  • DeepSeek V4 Pro allows ~11.7x more max output (384K vs 33K)
  • DeepSeek V4 Pro is ~4.0x cheaper on a blended token basis than GLM-5.2

Cost for 1M in + 250K out

Illustrative chat workload at primary-provider list prices.

DeepSeek V4 Pro
$0.65
GLM-5.2
$2.50

DeepSeek V4 Pro is about 3.8x cheaper on this workload.

Price per 1M tokens

Full comparison

Organization

DeepSeek

Zhipu AI

Family

DeepSeek

GLM

License

MIT

MIT

Open weights

Yes

Yes

Release

Apr 24, 2026

Jun 1, 2026

Knowledge cutoff

-

-

API / provider

DeepSeek

Zhipu / Z.ai

Modalities

text → text

text, image → text

Specs

Context window

1M

200K

A +800K

Max output

384K

33K

A +351K

Parameters

1.6T (49B act.)

—

—

Pricing

Input $/1M

$0.43

$1.40

A +$0.96

Output $/1M

$0.87

$4.40

A +$3.53

Blended $/1M (3∶1)

$0.54

$2.15

A +$1.61

Speed

tok/s

85

168

B +83

TTFT (s)

0.4

0.35

B +0.05

Reasoning

MMLU-Pro

—

—

—

GPQA Diamond

88.8

89.5

B +0.7 pts

Humanity's Last Exam

37.5

41.1

B +3.6 pts

AIME 2025

—

—

—

MATH-500

—

—

—

Humanity's Last Exam (with tools)

—

—

—

AA-Omniscience Accuracy

43.0

24.3

A +18.7 pts

AA-LCR v1.1

74.7

78.3

B +3.6 pts

CritPt

12.9

20.9

B +8.0 pts

MMMU-Pro

—

—

—

IFBench

76.5

73.3

A +3.2 pts

Chartography

—

—

—

Chartography (With Tools)

—

—

—

Coding

SWE-bench Verified

—

—

—

SWE-bench Pro

—

—

—

SWE-bench Multilingual

—

—

—

LiveCodeBench

—

—

—

Terminal-Bench 2.1

64.0

77.9

B +13.9 pts

Aider Polyglot

—

—

—

Terminal-Bench 3

—

4.6

—

BigCodeBench

—

—

—

SciCode

50.8

51.2

B +0.4 pts

CursorBench

—

55.0

—

SWE-Rebench

41.4

57.0

B +15.6 pts

NL2Repo-Bench

38.5

—

—

DeepSWE

12.8

44.0

B +31.2 pts

WebDev Arena

1,446

1,592

B +146 Elo

Terminal-Bench 4.0

—

—

—

CursorBench 4.0

—

—

—

FrontierCode v1.1 (Main)

—

—

—

Arena

LMArena Elo

1,457

1,472

B +15 Elo

DeepSeek V4 Pro

Context
1M / 384K out
Parameters
1.6T (49B act.)
Price
$0.43 / $0.87
Speed
85 tok/s · 0.4s TTFT
Modalities
text → text
License
MIT

Open-weight frontier MoE - exceptional quality-per-dollar with MIT license and 1M context. Scores shown are the original preview; the GA 0813 release improves agentic coding substantially.

GLM-5.2

Context
200K / 33K out
Parameters
—
Price
$1.40 / $4.40
Speed
168 tok/s · 0.35s TTFT
Modalities
text, image → text
License
MIT

Open MIT GLM-5.2 - competitive open Arena Elo and strong Chinese/English bilingual performance.

Benchmark charts

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

Reasoning & knowledge

GPQA: B +0.7HLE: B +3.6CritPt: B +8.0AA-LCR: B +3.6Omniscience Acc.: A +18.7IFBench: A +3.2

Coding

TermBench: B +13.9SciCode: B +0.4SWE-Rebench: B +15.6DeepSWE: B +31.2WebDev Elo: B +146

Tool use & function calling

Toolathlon V.: B +4.0τ³-Banking: B +4.5

Arena

Arena Elo: B +15

Composite indices

AA Index: B +3.1AA Index v4.3: B +3.1

Pick a different pair · Back to catalog