GLM-5.3 vs Kimi K3

At a glance

• Capability index

  • GLM-5.3155.6
  • Kimi K3157.4

  Kimi K3 is 1.8 points higher.

• Price per 1M tokens

  • GLM-5.3$2.15
  • Kimi K3$6.00

  GLM-5.3 is 2.8 times cheaper.

• Context window

  • GLM-5.31M tokens
  • Kimi K31M tokens

  They take in about the same.

• Released

  • GLM-5.3Aug 14, 2026
  • Kimi K3Jul 16, 2026

  GLM-5.3 is 29 days 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 GLM-5.3 vs Kimi K3

Measure	GLM-5.3Z.ai	Kimi K3Moonshot AI
Model
Maker	Z.ai	Moonshot AI
Sold here by	Z.AI	Moonshot AI
API model ID	glm-5.3	kimi-k3
Released	Aug 14, 2026	Jul 16, 2026
Knowledge cutoff	Not reported	Not reported
Weights	Open: downloadable	Open: downloadable
Developer’s country	China	China
Hosts selling it	75including Z.ai directly	84including Moonshot AI directly
Training compute	Not published	2.0 × 10²⁵ FLOPEpoch AI estimate
Test results
Capability index	155.6#27 of 274	157.4 (best of these models)#15 of 274
LiveBench overall	76.1%#32 of 66	79.2% (best of these models)#15 of 66
Arena text	1479#27 of 413	1488 (best of these models)#16 of 413
LiveBench coding	79.0%#26 of 66	81.5% (best of these models)#15 of 66
Arena WebDev	1623#20 of 138	1658 (best of these models)#13 of 138
Arena coding	1521#34 of 408	1541 (best of these models)#9 of 408
LiveBench mathematics	87.9% (best of these models)#43 of 66	84.4%#53 of 66
Mock AIME 2024–2025	91.1%#65 of 297	97.2% (best of these models)#31 of 297
FrontierMath Tiers 1–3	68.8%#25 of 114	72.2% (best of these models)#22 of 114
FrontierMath Tier 4	29.3%#36 of 70	39.0% (best of these models)#29 of 70
Arena math	1494#20 of 396	1501 (best of these models)#16 of 396
SimpleQA Verified	41.0%#44 of 86	50.6% (best of these models)#24 of 86
LiveBench agentic coding	60.9%#15 of 66	62.2% (best of these models)#10 of 66
DeepSWE	69.0% (best of these models)#16 of 69	68.5%#19 of 69
CursorBench	42.6%#20 of 62	Not tested
LiveBench language	79.9%#31 of 66	85.5% (best of these models)#11 of 66
LiveBench instructions	69.3%#33 of 66	71.4% (best of these models)#23 of 66
Arena creative writing	1454#34 of 411	1460 (best of these models)#26 of 411
Arena instruction following	1476#25 of 413	1488 (best of these models)#14 of 413
LiveBench data analysis	70.2%#51 of 66	78.7% (best of these models)#21 of 66
LiveBench reasoning	85.8%#35 of 66	90.7% (best of these models)#9 of 66
GPQA Diamond	90.9%#34 of 319	93.1% (best of these models)#20 of 319
Chess puzzles	21.0%#80 of 227	39.0% (best of these models)#31 of 227
Arena hard prompts	1503#25 of 413	1517 (best of these models)#9 of 413
Price per million tokens
Input	$1.40 (best of these models)	$3.00
Output	$4.40 (best of these models)	$15.00
Blended, 3 input to 1 output	$2.15 (best of these models)	$6.00
Cached input	$0.26 (best of these models)	$0.30
Writing to the cache	Free	Not reported
The maker’s own price	This listing	This listing
Limits
Context window	1M tokens	1M tokens
Max input	1M tokens	1M tokens
Max output	131.1K tokens	1M tokens (best of these models)
Features
Reads	text	images, text and video
Produces	text	text
Reasoning	Yeseffort low, high, max; can be switched off; thinking budget can be set	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	No	Yes
Temperature setting	Supported	Fixed by the provider

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.