Gemini 3.7 Flash vs GLM-5.3

At a glance

• Capability index

  • Gemini 3.7 Flash157.3
  • GLM-5.3155.6

  Gemini 3.7 Flash is 1.7 points higher.

• Price per 1M tokens

  • Gemini 3.7 Flash$1.50
  • GLM-5.3$2.15

  Gemini 3.7 Flash is 30% cheaper.

• Context window

  • Gemini 3.7 Flash1M tokens
  • GLM-5.31M tokens

  They take in about the same.

• Released

  • Gemini 3.7 FlashAug 13, 2026
  • GLM-5.3Aug 14, 2026

  GLM-5.3 is 1 day newer.

Where each scores higher

On the 22 tests both have taken, each on its own scale. A gap of a point or two can sit within a test’s margin of error.

1 more test has results for only one of them; the table below lists every result.

Everything side by side

Comparison of Gemini 3.7 Flash vs GLM-5.3

Measure	Gemini 3.7 FlashGoogle	GLM-5.3Z.ai
Model
Maker	Google	Z.ai
Sold here by	Google	Z.AI
API model ID	gemini-3.7-flash	glm-5.3
Released	Aug 13, 2026	Aug 14, 2026
Knowledge cutoff	Mar 2026	Not reported
Weights	Closed: hosted access only	Open: downloadable
Developer’s country	United States	China
Hosts selling it	29including Google directly	75including Z.ai directly
Test results
Capability index	157.3 (best of these models)#16 of 274	155.6#27 of 274
LiveBench overall	78.8% (best of these models)#16 of 66	76.1%#32 of 66
Arena text	1488 (best of these models)#17 of 413	1479#27 of 413
LiveBench coding	78.9%#29 of 66	79.0% (best of these models)#26 of 66
Arena WebDev	1592#27 of 138	1623 (best of these models)#20 of 138
Arena coding	1518#41 of 408	1521 (best of these models)#34 of 408
LiveBench mathematics	93.5% (best of these models)#20 of 66	87.9%#43 of 66
Mock AIME 2024–2025	97.2% (best of these models)#31 of 297	91.1%#65 of 297
FrontierMath Tiers 1–3	71.6% (best of these models)#23 of 114	68.8%#25 of 114
FrontierMath Tier 4	36.6% (best of these models)#30 of 70	29.3%#36 of 70
Arena math	1503 (best of these models)#12 of 396	1494#20 of 396
SimpleQA Verified	69.2% (best of these models)#9 of 86	41.0%#44 of 86
LiveBench agentic coding	58.3%#19 of 66	60.9% (best of these models)#15 of 66
DeepSWE	65.5%#25 of 69	69.0% (best of these models)#16 of 69
CursorBench	Not tested	42.6%#20 of 62
LiveBench language	85.5% (best of these models)#13 of 66	79.9%#31 of 66
LiveBench instructions	79.9% (best of these models)#2 of 66	69.3%#33 of 66
Arena creative writing	1493 (best of these models)#5 of 411	1454#34 of 411
Arena instruction following	1485 (best of these models)#17 of 413	1476#25 of 413
LiveBench data analysis	68.0%#55 of 66	70.2% (best of these models)#51 of 66
LiveBench reasoning	87.8% (best of these models)#25 of 66	85.8%#35 of 66
GPQA Diamond	94.8% (best of these models)#5 of 319	90.9%#34 of 319
Chess puzzles	47.0% (best of these models)#14 of 227	21.0%#80 of 227
Arena hard prompts	1508 (best of these models)#19 of 413	1503#25 of 413
Price per million tokens
Input	$0.75 (best of these models)	$1.40
Output	$3.75 (best of these models)	$4.40
Blended, 3 input to 1 output	$1.50 (best of these models)	$2.15
Cached input	$0.075 (best of these models)	$0.26
Writing to the cache	Not reported	Free
Audio input	$0.75	Not reported
The maker’s own price	This listing	This listing
Limits
Context window	1M tokens	1M tokens
Max input	936K tokens	1M tokens (best of these models)
Max output	65.5K tokens	131.1K tokens (best of these models)
Features
Reads	audio, images, PDFs, text and video	text
Produces	text	text
Reasoning	Yeseffort low, medium, high	Yeseffort low, high, max; can be switched off; thinking budget can be set
Tool calling	Yes	Yes
Structured output	Yes, such as JSON	Yes, such as JSON
File attachments	Yes	No
Temperature setting	Supported	Supported

Bold marks the better value in each row. Prices are each model’s own list price, or the middle price across hosts when the maker does not sell it, blended as three input tokens for every output token. Test results as published by LiveBench (Apache-2.0), Epoch AI (CC BY 4.0), Datacurve DeepSWE, compiled by Epoch AI (CC BY 4.0), Cursor CursorBench, compiled by Epoch AI (CC BY 4.0), Arena (CC BY 4.0); model details and prices from models.dev (MIT). Confirm pricing with the provider before use.